2026-06-12 13:49:18 +00:00
{
"cells": [
{
"cell_type": "markdown",
"id": "bea416d7",
"metadata": {},
"source": [
"# ZeaVis Edu — Corn Leaf Disease Classifier v3.0\n",
"\n",
"Mengklasifikasikan penyakit daun jagung (Bercak Daun, Hawar Daun, Karat Daun, Daun Sehat)\n",
"menggunakan EfficientNetV2B0 dengan **CBAM spatial attention**, **RandAugment + weather simulation**,\n",
"dan **temperature-scaled confidence calibration** untuk deployment real-world.\n",
"\n",
"Fokus v3.0: **robustness dunia nyata** — berbagai pencahayaan, resolusi, angle, dan background.\n"
]
},
{
"cell_type": "markdown",
"id": "aba4b688",
"metadata": {},
"source": [
"## 1. Persiapan Lingkungan\n",
"\n",
"Mengimpor pustaka, mengatur seed, dan mengoptimalkan konfigurasi.\n",
"**Presisi float32**, resolusi target **224×
"Augmentasi real-world via RandAugment pool 15 transformasi.\n",
"Confidence calibration via temperature scaling.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 1,
2026-06-12 13:49:18 +00:00
"id": "9dc08169",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Requirement already satisfied: tensorflow==2.19.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 1)) (2.19.0)\n",
"Requirement already satisfied: tensorflowjs==4.22.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 2)) (4.22.0)\n",
"Requirement already satisfied: gdown in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 3)) (6.1.0)\n",
"Requirement already satisfied: kagglehub in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 4)) (1.0.2)\n",
"Requirement already satisfied: numpy in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 5)) (2.1.3)\n",
"Requirement already satisfied: matplotlib in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 6)) (3.10.9)\n",
"Requirement already satisfied: seaborn in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 7)) (0.13.2)\n",
"Requirement already satisfied: pillow in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 8)) (12.2.0)\n",
"Requirement already satisfied: split-folders in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 9)) (0.6.1)\n",
"Requirement already satisfied: scikit-learn in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 10)) (1.9.0)\n",
"Requirement already satisfied: tf2onnx in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 11)) (1.17.0)\n",
"Requirement already satisfied: onnxruntime in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 12)) (1.26.0)\n",
"Requirement already satisfied: huggingface_hub in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 13)) (1.19.0)\n",
"Requirement already satisfied: absl-py>=1.0.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (2.4.0)\n",
"Requirement already satisfied: astunparse>=1.6.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (1.6.3)\n",
"Requirement already satisfied: flatbuffers>=24.3.25 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (25.12.19)\n",
"Requirement already satisfied: gast!=0.5.0,!=0.5.1,!=0.5.2,>=0.2.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (0.7.0)\n",
"Requirement already satisfied: google-pasta>=0.1.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (0.2.0)\n",
"Requirement already satisfied: libclang>=13.0.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (18.1.1)\n",
"Requirement already satisfied: opt-einsum>=2.3.2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (3.4.0)\n",
"Requirement already satisfied: packaging in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (23.2)\n",
"Requirement already satisfied: protobuf!=4.21.0,!=4.21.1,!=4.21.2,!=4.21.3,!=4.21.4,!=4.21.5,<6.0.0dev,>=3.20.3 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (5.29.6)\n",
"Requirement already satisfied: requests<3,>=2.21.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (2.34.2)\n",
"Requirement already satisfied: setuptools in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (82.0.1)\n",
"Requirement already satisfied: six>=1.12.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (1.17.0)\n",
"Requirement already satisfied: termcolor>=1.1.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (3.3.0)\n",
"Requirement already satisfied: typing-extensions>=3.6.6 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (4.15.0)\n",
"Requirement already satisfied: wrapt>=1.11.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (2.2.1)\n",
"Requirement already satisfied: grpcio<2.0,>=1.24.3 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (1.81.1)\n",
"Requirement already satisfied: tensorboard~=2.19.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (2.19.0)\n",
"Requirement already satisfied: keras>=3.5.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (3.14.1)\n",
"Requirement already satisfied: h5py>=3.11.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (3.14.0)\n",
"Requirement already satisfied: ml-dtypes<1.0.0,>=0.5.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow==2.19.0->-r requirements.txt (line 1)) (0.5.4)\n",
"Requirement already satisfied: flax>=0.7.2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.12.7)\n",
"Requirement already satisfied: importlib_resources>=5.9.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (7.1.0)\n",
"Requirement already satisfied: jax>=0.4.13 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.10.1)\n",
"Requirement already satisfied: jaxlib>=0.4.13 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.10.1)\n",
"Requirement already satisfied: tf-keras>=2.13.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (2.19.0)\n",
"Requirement already satisfied: tensorflow-decision-forests>=1.5.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (1.12.0)\n",
"Requirement already satisfied: tensorflow-hub>=0.16.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.16.1)\n",
"Requirement already satisfied: charset_normalizer<4,>=2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from requests<3,>=2.21.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (3.4.7)\n",
"Requirement already satisfied: idna<4,>=2.5 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from requests<3,>=2.21.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (3.18)\n",
"Requirement already satisfied: urllib3<3,>=1.26 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from requests<3,>=2.21.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (2.7.0)\n",
"Requirement already satisfied: certifi>=2023.5.7 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from requests<3,>=2.21.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (2026.5.20)\n",
"Requirement already satisfied: markdown>=2.6.8 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorboard~=2.19.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (3.10.2)\n",
"Requirement already satisfied: tensorboard-data-server<0.8.0,>=0.7.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorboard~=2.19.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (0.7.2)\n",
"Requirement already satisfied: werkzeug>=1.0.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorboard~=2.19.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (3.1.8)\n",
"Requirement already satisfied: beautifulsoup4 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from gdown->-r requirements.txt (line 3)) (4.15.0)\n",
"Requirement already satisfied: filelock in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from gdown->-r requirements.txt (line 3)) (3.29.3)\n",
"Requirement already satisfied: tqdm in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from gdown->-r requirements.txt (line 3)) (4.68.2)\n",
"Requirement already satisfied: kagglesdk<1.0,>=0.1.22 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from kagglehub->-r requirements.txt (line 4)) (0.1.29)\n",
"Requirement already satisfied: pyyaml in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from kagglehub->-r requirements.txt (line 4)) (6.0.3)\n",
"Requirement already satisfied: contourpy>=1.0.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (1.3.3)\n",
"Requirement already satisfied: cycler>=0.10 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (0.12.1)\n",
"Requirement already satisfied: fonttools>=4.22.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (4.63.0)\n",
"Requirement already satisfied: kiwisolver>=1.3.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (1.5.0)\n",
"Requirement already satisfied: pyparsing>=3 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (3.3.2)\n",
"Requirement already satisfied: python-dateutil>=2.7 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 6)) (2.9.0.post0)\n",
"Requirement already satisfied: pandas>=1.2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from seaborn->-r requirements.txt (line 7)) (3.0.3)\n",
"Requirement already satisfied: scipy>=1.10.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from scikit-learn->-r requirements.txt (line 10)) (1.17.1)\n",
"Requirement already satisfied: joblib>=1.4.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from scikit-learn->-r requirements.txt (line 10)) (1.5.3)\n",
"Requirement already satisfied: narwhals>=2.0.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from scikit-learn->-r requirements.txt (line 10)) (2.22.1)\n",
"Requirement already satisfied: threadpoolctl>=3.5.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from scikit-learn->-r requirements.txt (line 10)) (3.6.0)\n",
"Requirement already satisfied: onnx>=1.14.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tf2onnx->-r requirements.txt (line 11)) (1.21.0)\n",
"Requirement already satisfied: click>=8.4.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from huggingface_hub->-r requirements.txt (line 13)) (8.4.1)\n",
"Requirement already satisfied: fsspec>=2023.5.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from huggingface_hub->-r requirements.txt (line 13)) (2026.4.0)\n",
"Requirement already satisfied: hf-xet<2.0.0,>=1.5.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from huggingface_hub->-r requirements.txt (line 13)) (1.5.1)\n",
"Requirement already satisfied: httpx<1,>=0.23.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from huggingface_hub->-r requirements.txt (line 13)) (0.28.1)\n",
"Requirement already satisfied: typer<0.26.0,>=0.20.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from huggingface_hub->-r requirements.txt (line 13)) (0.25.1)\n",
"Requirement already satisfied: anyio in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from httpx<1,>=0.23.0->huggingface_hub->-r requirements.txt (line 13)) (4.13.0)\n",
"Requirement already satisfied: httpcore==1.* in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from httpx<1,>=0.23.0->huggingface_hub->-r requirements.txt (line 13)) (1.0.9)\n",
"Requirement already satisfied: h11>=0.16 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from httpcore==1.*->httpx<1,>=0.23.0->huggingface_hub->-r requirements.txt (line 13)) (0.16.0)\n",
"Requirement already satisfied: shellingham>=1.3.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (1.5.4)\n",
"Requirement already satisfied: rich>=13.8.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (15.0.0)\n",
"Requirement already satisfied: annotated-doc>=0.0.2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (0.0.4)\n",
"Requirement already satisfied: wheel<1.0,>=0.23.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from astunparse>=1.6.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (0.45.1)\n",
"Requirement already satisfied: msgpack in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (1.2.0)\n",
"Requirement already satisfied: optax in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.2.8)\n",
"Requirement already satisfied: orbax-checkpoint in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.12.0)\n",
"Requirement already satisfied: tensorstore in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.1.84)\n",
"Requirement already satisfied: treescope>=0.1.7 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.1.10)\n",
"Requirement already satisfied: namex in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from keras>=3.5.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (0.1.0)\n",
"Requirement already satisfied: optree in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from keras>=3.5.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (0.19.1)\n",
"Requirement already satisfied: markdown-it-py>=2.2.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from rich>=13.8.0->typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (4.2.0)\n",
"Requirement already satisfied: pygments<3.0.0,>=2.13.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from rich>=13.8.0->typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (2.20.0)\n",
"Requirement already satisfied: mdurl~=0.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from markdown-it-py>=2.2.0->rich>=13.8.0->typer<0.26.0,>=0.20.0->huggingface_hub->-r requirements.txt (line 13)) (0.1.2)\n",
"Requirement already satisfied: wurlitzer in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow-decision-forests>=1.5.0->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (3.1.1)\n",
"Requirement already satisfied: ydf>=0.11.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tensorflow-decision-forests>=1.5.0->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.16.0)\n",
"Requirement already satisfied: markupsafe>=2.1.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from werkzeug>=1.0.1->tensorboard~=2.19.0->tensorflow==2.19.0->-r requirements.txt (line 1)) (3.0.3)\n",
"Requirement already satisfied: soupsieve>=1.6.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from beautifulsoup4->gdown->-r requirements.txt (line 3)) (2.8.4)\n",
"Requirement already satisfied: etils[epath,epy] in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (1.14.0)\n",
"Requirement already satisfied: prometheus-client>=0.20.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.25.0)\n",
"Requirement already satisfied: aiofiles in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (25.1.0)\n",
"Requirement already satisfied: humanize in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (4.15.0)\n",
"Requirement already satisfied: simplejson>=3.16.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (4.1.1)\n",
"Requirement already satisfied: psutil in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (7.2.2)\n",
"Requirement already satisfied: uvloop in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.22.1)\n",
"Requirement already satisfied: zipp in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from etils[epath,epy]->orbax-checkpoint->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (4.1.0)\n",
"Requirement already satisfied: PySocks!=1.5.7,>=1.5.6 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from requests[socks]->gdown->-r requirements.txt (line 3)) (1.7.1)\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"!pip install -r requirements.txt\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 2,
2026-06-12 13:49:18 +00:00
"id": "9d108f5e",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:16:23.410032: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
"E0000 00:00:1781273783.433173 1198901 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
"E0000 00:00:1781273783.440756 1198901 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
"W0000 00:00:1781273783.459729 1198901 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\n",
"W0000 00:00:1781273783.459748 1198901 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\n",
"W0000 00:00:1781273783.459750 1198901 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\n",
"W0000 00:00:1781273783.459752 1198901 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\n",
"2026-06-12 14:16:23.465375: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
"To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Running locally (TF 2.19.0, GPU: [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')])\n",
"Setup OK. IMG=(224, 224), BATCH=32\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"import os, shutil, zipfile, random, time, json\n",
"from collections import Counter\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"import seaborn as sns\n",
"from PIL import Image\n",
"\n",
"import tensorflow as tf\n",
"from tensorflow.keras import layers, models, callbacks\n",
"from tensorflow.keras.applications import EfficientNetV2B0\n",
"from tensorflow.keras.applications.efficientnet_v2 import preprocess_input\n",
"from tensorflow.keras.optimizers import AdamW\n",
"from tensorflow.keras.optimizers.schedules import CosineDecay\n",
"from sklearn.metrics import classification_report, confusion_matrix\n",
"from sklearn.utils.class_weight import compute_class_weight\n",
"from sklearn.model_selection import train_test_split\n",
"\n",
"# Optional: perceptual hashing for dedup (pip install imagehash)\n",
"try:\n",
" import imagehash\n",
" HAS_IMAGEHASH = True\n",
"except ImportError:\n",
" HAS_IMAGEHASH = False\n",
"\n",
"# Optional: scipy for temperature optimization\n",
"try:\n",
" from scipy.optimize import minimize_scalar\n",
" HAS_SCIPY = True\n",
"except ImportError:\n",
" HAS_SCIPY = False\n",
"\n",
"# Detect environment\n",
"try:\n",
" from google.colab import drive\n",
" IS_COLAB = True\n",
" print(\"Running on Google Colab\")\n",
"except ModuleNotFoundError:\n",
" IS_COLAB = False\n",
" print(f\"Running locally (TF {tf.__version__}, GPU: {tf.config.list_physical_devices('GPU')})\")\n",
"\n",
"tf.keras.mixed_precision.set_global_policy('float32')\n",
"\n",
"# Hyperparams\n",
"IMG_SIZE = (224, 224)\n",
"BATCH_SIZE = 32\n",
"SEED = 42\n",
"random.seed(SEED)\n",
"np.random.seed(SEED)\n",
"tf.random.set_seed(SEED)\n",
"\n",
"AUTOTUNE = tf.data.AUTOTUNE\n",
"print(f\"Setup OK. IMG={IMG_SIZE}, BATCH={BATCH_SIZE}\")\n"
]
},
{
"cell_type": "markdown",
"id": "06382ddf",
"metadata": {},
"source": [
"## 2. Download dan Ekstraksi Dataset\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 3,
2026-06-12 13:49:18 +00:00
"id": "26afde45",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Dataset ready.\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"if IS_COLAB:\n",
" drive.mount('/content/drive')\n",
" archive_path = '/content/drive/MyDrive/jagung/dataset.zip'\n",
" destination_path = '/content/dataset.zip'\n",
" extract_path = '/content/dataset'\n",
"else:\n",
" base = os.getcwd()\n",
" archive_path = os.path.join(base, 'dataset.zip')\n",
" destination_path = archive_path\n",
" extract_path = os.path.join(base, 'dataset')\n",
"\n",
"if os.path.exists(destination_path):\n",
" if not os.path.exists(extract_path) or len(os.listdir(extract_path)) == 0:\n",
" os.makedirs(extract_path, exist_ok=True)\n",
" print(\"Extracting dataset...\")\n",
" try:\n",
" with zipfile.ZipFile(destination_path, 'r') as zip_ref:\n",
" zip_ref.extractall(path=extract_path)\n",
" print(\"Extraction completed!\")\n",
" except Exception as e:\n",
" print(f\"Extraction failed: {e}\")\n",
" else:\n",
" print(\"Dataset ready.\")\n",
"else:\n",
" print(f\"dataset.zip not found at {destination_path}. Upload dataset.zip to Google Drive / MyDrive/jagung/\")\n",
" print(\"Or run preprocessing.py locally and upload the resulting dataset.zip\")"
]
},
{
"cell_type": "markdown",
"id": "199bdf60",
"metadata": {},
"source": [
"## 3. Data Cleaning — Corrupt Detection + Augmented Dedup\n",
"\n",
"Membersihkan dataset dari:\n",
"- File corrupt / tidak bisa dibuka PIL\n",
"- File `augmented_*` (pre-augmented duplicates — menyebabkan data leakage)\n",
"- Gambar dengan dimensi atau aspect ratio ekstrim\n",
"- Gambar dengan variance terlalu rendah (hampir seragam)\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 4,
2026-06-12 13:49:18 +00:00
"id": "75cacdf5",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Dataset path: /home/asephs/ZeaVis-Edu/Machine_Learning/dataset\n",
"Split dataset already exists. Skipping validation and split.\n",
"Train: 3526 | Val: 754 | Test: 758\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"# --- Determine dataset path ---\n",
"dataset_path = extract_path\n",
"\n",
"print(f\"Dataset path: {dataset_path}\")\n",
"\n",
"# Check if dataset already validated and split exists — skip if so\n",
"split_output_dir = \"/content/dataset_split\" if IS_COLAB else os.path.join(os.getcwd(), \"dataset_split\")\n",
"if os.path.exists(os.path.join(split_output_dir, 'train')):\n",
" print(\"Split dataset already exists. Skipping validation and split.\")\n",
" # Still need class_names and counts for downstream cells\n",
" train_dir = os.path.join(split_output_dir, 'train')\n",
" val_dir = os.path.join(split_output_dir, 'val')\n",
" test_dir = os.path.join(split_output_dir, 'test')\n",
" print(f'Train: {sum(len(files) for _, _, files in os.walk(train_dir))} | '\n",
" f'Val: {sum(len(files) for _, _, files in os.walk(val_dir))} | '\n",
" f'Test: {sum(len(files) for _, _, files in os.walk(test_dir))}')\n",
"else:\n",
" MIN_FILE_SIZE = 512\n",
" MIN_DIM = 32\n",
" MAX_ASPECT = 5.0\n",
"\n",
" def remove_augmented_duplicates(directory):\n",
" removed = 0\n",
" for root, dirs, files in os.walk(directory):\n",
" for file in files:\n",
" if file.startswith(\"augmented_\"):\n",
" try:\n",
" os.remove(os.path.join(root, file))\n",
" removed += 1\n",
" except OSError:\n",
" pass\n",
" return removed\n",
"\n",
" def clean_and_validate_images(directory):\n",
" stats = {\"too_small\": 0, \"corrupt\": 0, \"small_dims\": 0, \"extreme_aspect\": 0, \"low_var\": 0, \"ok\": 0}\n",
" for root, dirs, files in os.walk(directory):\n",
" for file in files:\n",
" fp = os.path.join(root, file)\n",
" try:\n",
" if os.path.getsize(fp) < MIN_FILE_SIZE:\n",
" os.remove(fp); stats[\"too_small\"] += 1; continue\n",
" except OSError:\n",
" continue\n",
" try:\n",
" img = Image.open(fp); img.verify()\n",
" except Exception:\n",
" try: os.remove(fp); stats[\"corrupt\"] += 1\n",
" except OSError: pass\n",
" continue\n",
" try:\n",
" img = Image.open(fp)\n",
" w, h = img.size\n",
" if w < MIN_DIM or h < MIN_DIM:\n",
" os.remove(fp); stats[\"small_dims\"] += 1; continue\n",
" aspect = w / max(h, 1)\n",
" if aspect > MAX_ASPECT or aspect < 1.0 / MAX_ASPECT:\n",
" os.remove(fp); stats[\"extreme_aspect\"] += 1; continue\n",
" if img.mode not in ('RGB', 'RGBA'):\n",
" img = img.convert('RGB'); img.save(fp)\n",
" arr = np.array(img).astype(np.float32)\n",
" if np.std(arr) < 2.0:\n",
" os.remove(fp); stats[\"low_var\"] += 1; continue\n",
" stats[\"ok\"] += 1\n",
" except Exception:\n",
" try: os.remove(fp); stats[\"corrupt\"] += 1\n",
" except OSError: pass\n",
" return stats\n",
"\n",
" print(\"1. Removing augmented duplicates...\")\n",
" n_aug = remove_augmented_duplicates(dataset_path)\n",
" print(f\" Removed {n_aug} augmented_* files\")\n",
"\n",
" print(\"2. Validating images...\")\n",
" stats = clean_and_validate_images(dataset_path)\n",
" print(f\" OK: {stats['ok']} | Removed: too_small={stats['too_small']} corrupt={stats['corrupt']} \"\n",
" f\"small_dims={stats['small_dims']} aspect={stats['extreme_aspect']} low_var={stats['low_var']}\")\n",
"\n",
" if HAS_IMAGEHASH:\n",
" print(\"3. Perceptual hash dedup...\")\n",
" seen, removed = {}, 0\n",
" for cn in sorted(os.listdir(dataset_path)):\n",
" cp = os.path.join(dataset_path, cn)\n",
" if not os.path.isdir(cp): continue\n",
" for f in sorted(os.listdir(cp)):\n",
" fp = os.path.join(cp, f)\n",
" if not os.path.isfile(fp): continue\n",
" try:\n",
" ah = imagehash.average_hash(Image.open(fp).convert('RGB'))\n",
" for sk, (sp, sc) in seen.items():\n",
" if ah - imagehash.hex_to_hash(sk) <= 5:\n",
" try: os.remove(fp); removed += 1\n",
" except OSError: pass\n",
" break\n",
" else:\n",
" seen[str(ah)] = (fp, cn)\n",
" except Exception:\n",
" pass\n",
" print(f\" Removed {removed} near-duplicates\")\n",
" else:\n",
" print(\"3. Perceptual hash dedup SKIPPED (pip install imagehash)\")\n",
"\n",
" total = sum(len(files) for _, _, files in os.walk(dataset_path))\n",
" print(f\"\\nTotal clean images: {total}\")"
]
},
{
"cell_type": "markdown",
"id": "fbe9498c",
"metadata": {},
"source": [
"## 4. Stratified Split by Source (70:15:15)\n",
"\n",
"**Tidak menggunakan splitfolders!** Split manual dengan stratifikasi berdasarkan prefix sumber gambar.\n",
"Ini mencegah gambar dari sesi foto yang sama (lighting & background identik) masuk ke train DAN test.\n",
"\n",
"Source prefixes:\n",
"- `IMG_*` → foto HP\n",
"- `Corn_*` → dataset lab publik\n",
"- `CBS*`, `GLS*`, `NLS*`, `CLS*` → berbagai dataset lab\n",
"- `SCR*`, `CR*`, `NLB*`, `SLB*` → dataset spesifik penyakit\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 5,
2026-06-12 13:49:18 +00:00
"id": "56234c24",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Split already exists. Skipping.\n",
"Train: 3526 | Val: 754 | Test: 758\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"def extract_source_prefix(filename):\n",
" f = os.path.splitext(filename)[0]\n",
" if f.startswith('IMG_'): return 'phone'\n",
" if f.startswith('Corn_'): return 'lab_corn'\n",
" for prefix in ['CBS', 'GLS', 'NLS', 'CLS']:\n",
" if f.startswith(prefix): return 'lab_disease'\n",
" for prefix in ['SCR', 'CR', 'NLB', 'SLB', 'SRS']:\n",
" if f.startswith(prefix): return 'lab_rust_blight'\n",
" return 'other'\n",
"\n",
"def stratified_split_by_source(dataset_path, output_dir, ratios=(0.7, 0.15, 0.15), seed=42):\n",
" class_images = {}\n",
" for cn in sorted(os.listdir(dataset_path)):\n",
" cp = os.path.join(dataset_path, cn)\n",
" if not os.path.isdir(cp): continue\n",
" class_images[cn] = []\n",
" for f in os.listdir(cp):\n",
" fp = os.path.join(cp, f)\n",
" if os.path.isfile(fp):\n",
" class_images[cn].append((fp, f, extract_source_prefix(f)))\n",
"\n",
" for split in ['train', 'val', 'test']:\n",
" for cn in class_images:\n",
" os.makedirs(os.path.join(output_dir, split, cn), exist_ok=True)\n",
"\n",
" rng = np.random.RandomState(seed)\n",
"\n",
" for cn, images in class_images.items():\n",
" by_source = {}\n",
" for fp, fn, src in images:\n",
" by_source.setdefault(src, []).append((fp, fn))\n",
"\n",
" train_files, val_files, test_files = [], [], []\n",
" for src, src_images in by_source.items():\n",
" n = len(src_images)\n",
" rng.shuffle(src_images)\n",
" n_train = max(1, int(n * ratios[0]))\n",
" n_val = max(1, int(n * ratios[1]))\n",
" train_files.extend(src_images[:n_train])\n",
" val_files.extend(src_images[n_train:n_train + n_val])\n",
" test_files.extend(src_images[n_train + n_val:])\n",
"\n",
" for fp, fn in train_files:\n",
" shutil.copy(fp, os.path.join(output_dir, 'train', cn, fn))\n",
" for fp, fn in val_files:\n",
" shutil.copy(fp, os.path.join(output_dir, 'val', cn, fn))\n",
" for fp, fn in test_files:\n",
" shutil.copy(fp, os.path.join(output_dir, 'test', cn, fn))\n",
"\n",
" train_srcs = Counter(extract_source_prefix(fn) for _, fn in train_files)\n",
" print(f\" {cn}: train={len(train_files)} val={len(val_files)} test={len(test_files)} | \"\n",
" f\"sources={dict(train_srcs)}\")\n",
"\n",
"output_dir = split_output_dir # Already defined in previous cell\n",
"\n",
"if os.path.exists(os.path.join(output_dir, 'train')):\n",
" print(\"Split already exists. Skipping.\")\n",
"else:\n",
" if os.path.exists(output_dir):\n",
" shutil.rmtree(output_dir)\n",
"\n",
" print(\"Splitting dataset 70:15:15 (stratified by source)...\")\n",
" stratified_split_by_source(dataset_path, output_dir, seed=SEED)\n",
" print(\"Done.\")\n",
"\n",
"train_dir = os.path.join(output_dir, 'train')\n",
"val_dir = os.path.join(output_dir, 'val')\n",
"test_dir = os.path.join(output_dir, 'test')\n",
"\n",
"def count_images(path):\n",
" return sum(len(files) for _, _, files in os.walk(path))\n",
"\n",
"print(f'Train: {count_images(train_dir)} | Val: {count_images(val_dir)} | Test: {count_images(test_dir)}')"
]
},
{
"cell_type": "markdown",
"id": "3eead964",
"metadata": {},
"source": [
"## 5. RandAugment Pipeline dengan Weather Simulation\n",
"\n",
"Pipeline augmentasi baru untuk **robustness dunia nyata**:\n",
"\n",
"### Pool 15 Transformasi (RandAugment: pilih N=3 per gambar)\n",
"| Kategori | Transformasi | Simulasi |\n",
"|---|---|---|\n",
"| Geometric | Flip, Rotate, Zoom, Translate, Shear | Variasi angle/jarak foto |\n",
"| Color/Light | Hue, Saturation, Brightness, Contrast, Solarize | Variasi kamera & waktu hari |\n",
"| Weather | Fog, Shadow | Kondisi lapangan berkabut/berbayang |\n",
"| Degradation | GaussianBlur, ResolutionDrop | Blur gerakan, kamera rendah |\n",
"| Mixing | MixUp, CutMix, RandomErasing | Regularisasi label & occlusions |\n",
"\n",
"Fog dan Shadow adalah **custom tf operations** — tidak ada di Keras layers standar.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 6,
2026-06-12 13:49:18 +00:00
"id": "8535054d",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 3526 files belonging to 4 classes.\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"I0000 00:00:1781273786.135481 1198901 gpu_device.cc:2019] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 37480 MB memory: -> device: 0, name: NVIDIA A100-SXM4-40GB, pci bus id: 0000:05:00.0, compute capability: 8.0\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 754 files belonging to 4 classes.\n",
"Found 758 files belonging to 4 classes.\n",
"Classes (4): ['Bercak Daun', 'Daun Sehat', 'Hawar Daun', 'Karat Daun']\n",
"WARNING:tensorflow:From /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages/tensorflow/python/util/deprecation.py:660: calling map_fn_v2 (from tensorflow.python.ops.map_fn) with dtype is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Use fn_output_signature instead\n",
"Class weights (capped at 3.0):\n",
" Bercak Daun: 2.1983\n",
" Daun Sehat: 0.5826\n",
" Hawar Daun: 1.2593\n",
" Karat Daun: 0.9666\n",
"Data pipelines ready.\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"# ─── RandAugment Utilities ───\n",
"\n",
"def sample_beta_distribution(size, a=0.2, b=0.2):\n",
" g1 = tf.random.gamma([size], a, dtype=tf.float32)\n",
" g2 = tf.random.gamma([size], b, dtype=tf.float32)\n",
" return g2 / (g1 + g2 + 1e-8)\n",
"\n",
"def _apply_contrast(img, factor):\n",
" mean = tf.reduce_mean(tf.cast(img, tf.float32), axis=(0, 1), keepdims=True)\n",
" return mean + factor * (tf.cast(img, tf.float32) - mean)\n",
"\n",
"def _apply_brightness(img, delta):\n",
" return tf.clip_by_value(tf.cast(img, tf.float32) + delta, 0.0, 255.0)\n",
"\n",
"def _apply_hue(img, delta):\n",
" hsv = tf.image.rgb_to_hsv(tf.cast(img, tf.float32) / 255.0)\n",
" h = hsv[..., 0] + delta\n",
" h = h - tf.floor(h)\n",
" hsv_h = tf.stack([h, hsv[..., 1], hsv[..., 2]], axis=-1)\n",
" return tf.image.hsv_to_rgb(hsv_h) * 255.0\n",
"\n",
"@tf.function(reduce_retracing=True)\n",
"def _randaug_select(images, ops_per_image=3, magnitude=0.7):\n",
" # Graph-mode-safe RandAugment. No tf.image.random_* or Keras layers\n",
" # (both trace as Python bool checks on tensor bounds internally).\n",
"\n",
" def _zoom(img):\n",
" h = tf.cast(tf.shape(img)[0], tf.float32)\n",
" w = tf.cast(tf.shape(img)[1], tf.float32)\n",
" z = tf.random.uniform([], 1.0 - 0.2 * magnitude, 1.0 + 0.2 * magnitude)\n",
" zh = tf.cast(h / z, tf.int32); zw = tf.cast(w / z, tf.int32)\n",
" zoomed = tf.image.resize(tf.expand_dims(img, 0), [zh, zw], method='bilinear')[0]\n",
" return tf.image.resize_with_crop_or_pad(zoomed, tf.cast(h, tf.int32), tf.cast(w, tf.int32))\n",
" def _translate(img):\n",
" h, w = tf.shape(img)[0], tf.shape(img)[1]\n",
" tx = tf.cast(tf.random.uniform([], -0.15 * magnitude, 0.15 * magnitude) * tf.cast(w, tf.float32), tf.int32)\n",
" ty = tf.cast(tf.random.uniform([], -0.15 * magnitude, 0.15 * magnitude) * tf.cast(h, tf.float32), tf.int32)\n",
" return tf.roll(img, [ty, tx], axis=[0, 1])\n",
"\n",
" def op_flip(img): return tf.image.random_flip_left_right(img)\n",
" def op_flip_v(img): return tf.image.random_flip_up_down(img)\n",
" def op_rotate_90(img):\n",
" k = tf.random.uniform([], 0, 4, dtype=tf.int32)\n",
" return tf.image.rot90(img, k)\n",
" def op_zoom(img): return _zoom(img)\n",
" def op_translate(img): return _translate(img)\n",
" def op_contrast(img):\n",
" f = tf.random.uniform([], 1.0 - 0.5 * magnitude, 1.0 + 0.5 * magnitude)\n",
" return _apply_contrast(img, f)\n",
" def op_brightness(img):\n",
" d = tf.random.uniform([], -0.3 * magnitude * 255.0, 0.3 * magnitude * 255.0)\n",
" return _apply_brightness(img, d)\n",
" def op_hue(img):\n",
" d = tf.random.uniform([], -0.08 * tf.maximum(magnitude, 0.01), 0.08 * tf.maximum(magnitude, 0.01))\n",
" return _apply_hue(img, d)\n",
" def op_saturation(img):\n",
" lo = tf.maximum(0.5, 1.0 - 0.8 * magnitude)\n",
" hi = 1.0 + 0.8 * magnitude\n",
" f = tf.random.uniform([], lo, hi)\n",
" hsv = tf.image.rgb_to_hsv(tf.cast(img, tf.float32) / 255.0)\n",
" s = tf.clip_by_value(hsv[..., 1] * f, 0.0, 1.0)\n",
" return tf.image.hsv_to_rgb(tf.stack([hsv[..., 0], s, hsv[..., 2]], axis=-1)) * 255.0\n",
" def op_solarize(img):\n",
" thresh = tf.random.uniform([], 0.3, 0.8) * 255.0\n",
" f = tf.cast(img, tf.float32)\n",
" return tf.where(f < thresh, f, 255.0 - f)\n",
" def op_blur(img):\n",
" h, w = tf.shape(img)[0], tf.shape(img)[1]\n",
" sf = tf.random.uniform([], 2, 4, dtype=tf.int32)\n",
" small = tf.image.resize(tf.expand_dims(tf.cast(img, tf.float32), 0), [h // sf, w // sf], method='bilinear')\n",
" return tf.image.resize(small, [h, w], method='bilinear')[0]\n",
" def op_fog(img):\n",
" fl = tf.random.uniform([], 0.1, 0.1 + 0.4 * magnitude)\n",
" fc = tf.random.uniform([3], 0.7, 1.0) * 255.0\n",
" return tf.cast(img, tf.float32) * (1.0 - fl) + tf.reshape(fc, [1, 1, 3]) * fl\n",
" def op_shadow(img):\n",
" op = tf.random.uniform([], 0.2, 0.2 + 0.5 * magnitude)\n",
" return tf.cast(img, tf.float32) * (1.0 - op * 0.6)\n",
" def op_resolution_drop(img):\n",
" h, w = tf.shape(img)[0], tf.shape(img)[1]\n",
" sf = tf.random.uniform([], 2, 5, dtype=tf.int32)\n",
" small = tf.image.resize(tf.expand_dims(tf.cast(img, tf.float32), 0), [h // sf, w // sf], method='bilinear')\n",
" return tf.image.resize(small, [h, w], method='nearest')[0]\n",
" def op_identity(img): return tf.cast(img, tf.float32)\n",
"\n",
" ops = [op_flip, op_flip_v, op_rotate_90, op_zoom, op_translate,\n",
" op_contrast, op_brightness, op_hue, op_saturation,\n",
" op_solarize, op_blur, op_fog, op_shadow, op_resolution_drop, op_identity]\n",
"\n",
" def apply_randaug_single(img3d):\n",
" indices = tf.random.shuffle(tf.range(15))[:3] # ops_per_image=3, static\n",
" result = tf.cast(img3d, tf.float32)\n",
" # Unrolled static 3 iterations — avoids TF shape invariance error from tf.range loop\n",
" def _apply_one(r, idx):\n",
" return tf.switch_case(idx, {j: lambda j=j: ops[j](r) for j in range(15)})\n",
" i0, i1, i2 = indices[0], indices[1], indices[2]\n",
" result = _apply_one(result, i0)\n",
" result = _apply_one(result, i1)\n",
" result = _apply_one(result, i2)\n",
" return tf.clip_by_value(result, 0.0, 255.0)\n",
"\n",
" return tf.map_fn(apply_randaug_single, images, dtype=tf.float32, parallel_iterations=8)\n",
"\n",
"# Compatibility wrapper using Keras Sequential for basic geometric ops (kept for visualization)\n",
"geo_aug = tf.keras.Sequential([\n",
" layers.RandomFlip(\"horizontal_and_vertical\"),\n",
" layers.RandomRotation(0.15),\n",
" layers.RandomZoom(0.15),\n",
" layers.RandomTranslation(0.1, 0.1),\n",
" layers.RandomContrast(0.15),\n",
" layers.RandomBrightness(0.15),\n",
"], name=\"geo_aug\")\n",
"\n",
"# ─── MixUp & CutMix (unchanged from original) ───\n",
"\n",
"def mix_up(images, labels, alpha=0.2):\n",
" bs = tf.shape(images)[0]\n",
" lam = sample_beta_distribution(bs, alpha, alpha)\n",
" lam_img = tf.reshape(lam, [bs, 1, 1, 1])\n",
" ri = tf.random.shuffle(tf.range(bs))\n",
" mixed_img = lam_img * images + (1 - lam_img) * tf.gather(images, ri)\n",
" labels = tf.cast(labels, tf.float32)\n",
" lam_lbl = tf.reshape(lam, [-1, 1])\n",
" mixed_lbl = lam_lbl * labels + (1 - lam_lbl) * tf.gather(labels, ri)\n",
" return mixed_img, mixed_lbl\n",
"\n",
"def cut_mix(images, labels, alpha=0.2):\n",
" bs = tf.shape(images)[0]; h = tf.shape(images)[1]; w = tf.shape(images)[2]\n",
" lam = sample_beta_distribution(bs, alpha, alpha); ri = tf.random.shuffle(tf.range(bs))\n",
" cr = tf.sqrt(1.0 - lam)\n",
" rh = tf.cast(cr * tf.cast(h, tf.float32), tf.int32); rw = tf.cast(cr * tf.cast(w, tf.float32), tf.int32)\n",
" cx = tf.random.uniform([bs], 0, w, tf.int32); cy = tf.random.uniform([bs], 0, h, tf.int32)\n",
" hh = rh // 2; hw = rw // 2\n",
" x1 = tf.clip_by_value(cx - hw, 0, w); x2 = tf.clip_by_value(cx + hw, 0, w)\n",
" y1 = tf.clip_by_value(cy - hh, 0, h); y2 = tf.clip_by_value(cy + hh, 0, h)\n",
" col = tf.range(w, dtype=tf.int32); row = tf.range(h, dtype=tf.int32)\n",
" in_x = tf.logical_and(tf.reshape(col, [1, 1, w]) >= tf.reshape(x1, [bs, 1, 1]),\n",
" tf.reshape(col, [1, 1, w]) < tf.reshape(x2, [bs, 1, 1]))\n",
" in_y = tf.logical_and(tf.reshape(row, [1, h, 1]) >= tf.reshape(y1, [bs, 1, 1]),\n",
" tf.reshape(row, [1, h, 1]) < tf.reshape(y2, [bs, 1, 1]))\n",
" cm = tf.cast(tf.logical_and(in_y, in_x), tf.float32); cm = tf.expand_dims(cm, -1)\n",
" shuf = tf.gather(images, ri)\n",
" mi = (1.0 - cm) * images + cm * shuf\n",
" labels = tf.cast(labels, tf.float32); lr = tf.reshape(lam, [-1, 1])\n",
" ml = lr * labels + (1.0 - lr) * tf.gather(labels, ri)\n",
" return mi, ml\n",
"\n",
"def random_erasing(images, probability=0.25, scale=(0.02, 0.25)):\n",
" bs = tf.shape(images)[0]; h = tf.shape(images)[1]; w = tf.shape(images)[2]\n",
" ta = tf.random.uniform([], scale[0], scale[1]) * tf.cast(h * w, tf.float32)\n",
" ar = tf.random.uniform([], 0.3, 3.3)\n",
" eh = tf.cast(tf.math.sqrt(ta / ar), tf.int32); ew = tf.cast(tf.math.sqrt(ta * ar), tf.int32)\n",
" eh = tf.clip_by_value(eh, 1, h - 1); ew = tf.clip_by_value(ew, 1, w - 1)\n",
" cx = tf.random.uniform([], 0, w - ew, tf.int32); cy = tf.random.uniform([], 0, h - eh, tf.int32)\n",
" col = tf.range(w, dtype=tf.int32); row = tf.range(h, dtype=tf.int32)\n",
" ix = tf.logical_and(col >= cx, col < cx + ew)\n",
" iy = tf.logical_and(row >= cy, row < cy + eh)\n",
" em = tf.cast(tf.expand_dims(iy, 1) & tf.expand_dims(ix, 0), tf.float32)\n",
" em = tf.expand_dims(tf.expand_dims(em, 0), -1)\n",
" noise = tf.random.uniform([bs, eh, ew, 3], 0.0, 255.0, dtype=tf.float32)\n",
" pads = [[0, 0], [cy, h - (cy + eh)], [cx, w - (cx + ew)], [0, 0]]\n",
" npad = tf.pad(noise, pads, constant_values=0.0)\n",
" erased = images * (1.0 - em) + npad * em\n",
" return tf.cond(tf.random.uniform([]) < probability, lambda: erased, lambda: images)\n",
"\n",
"# ─── Main augmentation pipeline ───\n",
"\n",
"def augment_and_mix(images, labels):\n",
" # 1. RandAugment (geometric + color + weather + degradation)\n",
" images = tf.cast(images, tf.float32)\n",
" images = _randaug_select(images, ops_per_image=3, magnitude=0.7)\n",
" # 2. MixUp or CutMix (40% chance total: 20% MixUp, 20% CutMix)\n",
" choice = tf.random.uniform([])\n",
" labels_oh = tf.one_hot(labels, NUM_CLASSES)\n",
" images, labels_oh = tf.cond(\n",
" choice < 0.2, lambda: mix_up(images, labels_oh),\n",
" lambda: tf.cond(choice < 0.4, lambda: cut_mix(images, labels_oh),\n",
" lambda: (images, labels_oh)))\n",
" # 3. Random Erasing\n",
" images = random_erasing(images, probability=0.2)\n",
" return images, labels_oh\n",
"\n",
"# ─── Preprocessing ───\n",
"def preprocess_fn(image, label):\n",
" return preprocess_input(image), label\n",
"\n",
"# ─── Class weights ───\n",
"train_class_counts = Counter()\n",
"for cn in sorted(os.listdir(train_dir)):\n",
" p = os.path.join(train_dir, cn)\n",
" if os.path.isdir(p):\n",
" train_class_counts[cn] = len(os.listdir(p))\n",
"\n",
"y_int = []\n",
"for i, cn in enumerate(sorted(os.listdir(train_dir))):\n",
" cp = os.path.join(train_dir, cn)\n",
" if os.path.isdir(cp):\n",
" y_int.extend([i] * len(os.listdir(cp)))\n",
"\n",
"cw_array = compute_class_weight('balanced', classes=np.unique(y_int), y=y_int)\n",
"cw_capped = [min(w, 3.0) for w in cw_array]\n",
"class_weights_tensor = tf.constant(cw_capped, dtype=tf.float32)\n",
"\n",
"def add_sample_weight(image, label):\n",
" ci = tf.argmax(label, axis=-1)\n",
" sw = tf.gather(class_weights_tensor, ci)\n",
" return image, label, sw\n",
"\n",
"# ─── Build datasets ───\n",
"train_ds = tf.keras.utils.image_dataset_from_directory(\n",
" train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)\n",
"val_ds = tf.keras.utils.image_dataset_from_directory(\n",
" val_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)\n",
"test_ds = tf.keras.utils.image_dataset_from_directory(\n",
" test_dir, shuffle=False, batch_size=BATCH_SIZE, image_size=IMG_SIZE)\n",
"\n",
"class_names = train_ds.class_names\n",
"NUM_CLASSES = len(class_names)\n",
"print(f\"Classes ({NUM_CLASSES}): {class_names}\")\n",
"\n",
"train_ds = (train_ds\n",
" .map(augment_and_mix, num_parallel_calls=AUTOTUNE)\n",
" .map(preprocess_fn, num_parallel_calls=AUTOTUNE)\n",
" .map(add_sample_weight, num_parallel_calls=AUTOTUNE)\n",
" .prefetch(AUTOTUNE))\n",
"\n",
"val_ds = (val_ds\n",
" .map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)\n",
" .map(lambda img, lbl: (img, tf.one_hot(lbl, NUM_CLASSES)), num_parallel_calls=AUTOTUNE)\n",
" .prefetch(AUTOTUNE))\n",
"\n",
"test_ds = (test_ds\n",
" .map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)\n",
" .map(lambda img, lbl: (img, tf.one_hot(lbl, NUM_CLASSES)), num_parallel_calls=AUTOTUNE)\n",
" .prefetch(AUTOTUNE))\n",
"\n",
"print(\"Class weights (capped at 3.0):\")\n",
"for i, cn in enumerate(class_names):\n",
" if i < len(cw_capped):\n",
" print(f\" {cn}: {cw_capped[i]:.4f}\")\n",
"print(\"Data pipelines ready.\")\n"
]
},
{
"cell_type": "markdown",
"id": "4ca2ec40",
"metadata": {},
"source": [
"## 6. Visualisasi Sampel Data (Augmentasi Real-World)\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 7,
2026-06-12 13:49:18 +00:00
"id": "44d6d2b0",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 3526 files belonging to 4 classes.\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:16:33.570217: I tensorflow/core/framework/local_rendezvous.cc:407] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n"
]
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAABbEAAAPZCAYAAAAvMWctAAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjEwLjksIGh0dHBzOi8vbWF0cGxvdGxpYi5vcmcvJkbTWQAAAAlwSFlzAAAPYQAAD2EBqD+naQABAABJREFUeJzsvXm8JEWV9/2NiMyquvd209Dsy7AIAoMbyKKA0AiCyCaKioLSCoziDKgvjsOM4ygzuD24jdsDjoCAIggIiCzKIwIugMOiOC7gxjqgAt30cpeqzIh4/zgRkVl9b7cNNNJo/Pxgd1dlZUZmRpw453c25b33ZGRkZGRkZGRkZGRkZGRkZGRkZGRkZKyG0E/3ADIyMjIyMjIyMjIyMjIyMjIyMjIyMjKWh0xiZ2RkZGRkZGRkZGRkZGRkZGRkZGRkrLbIJHZGRkZGRkZGRkZGRkZGRkZGRkZGRsZqi0xiZ2RkZGRkZGRkZGRkZGRkZGRkZGRkrLbIJHZGRkZGRkZGRkZGRkZGRkZGRkZGRsZqi0xiZ2RkZGRkZGRkZGRkZGRkZGRkZGRkrLbIJHZGRkZGRkZGRkZGRkZGRkZGRkZGRsZqi0xiZ2RkZGRkZGRkZGRkZGRkZGRkZGRkrLbIJHZGRkZGRkZGRkZGRkZGRkZGRkZGRsZqi0xiZ2RkZGRkrAQ233xzlFJD/3W7XTbZZBNe+cpXcsUVVzzdQ1wpXH/99Sil2GuvvVZ43B/+8Ac6nQ5KKXbcccc/z+AyVku8+c1vTnN+++23X+Gxt9xyy9Aa+cEPfjD0/cknn4xSipNPPvlJj+vss89GKcXmm2++wuPuueeeNJ577rnnSV93dUT7Htv/jY6O8uxnP5tjjjmG//mf/3m6h/m43/+1116LUoo111wTa+2Mx3zwgx9c7nyLuPXWW5PMnpycfKLDXynEefnmN7/5cf0uvsM/NZ9nwq9//WuOP/54tttuO8bGxuj1emyyySbsvPPOHH/88Xz961+f9pu99toLpRTXX3/9477enxN/7nE+mfeQkZGRkZGR8dQik9gZGRkZGRmPA7vvvjvz589n/vz5HHDAARRFweWXX87BBx/MiSee+HQPb5Xh3HPPpaoqAG6//XbuuOOOp3lEf3lYWYfC6oQ77riD2267bbnfn3nmmX/G0WTMhMMOOyzJqD322INHHnmEs846ixe+8IVccsklT/fwHhd22203Op0OixYt4sc//vGMx1x33XXp78sjOuMxL3rRixgZGVnl43w6cckll/C85z2Pz3/+8/zxj39k991357DDDuP5z38+//u//8vnP/953va2tz3dw1xtEB3Sf6kOrYyMjIyMjL9kZBI7IyMjIyPjceDYY4/l7LPP5uyzz+bSSy/lN7/5DccffzwAn/rUp7jlllue5hGuGpx11lkAbLzxxkAmJzNgp512Apq5sSwmJye54IIL2HDDDdlkk01mPOb444/nl7/8ZVozGaseH//4x5OM+va3v81vfvMbdtttN+q65m1ve9tTHom8KjEyMsKLX/xiYJisjhgMBtx0000897nPpdPpLJfEjp+/9KUvfaqG+rTgD3/4A/Pnz6ff7/Pud7+bBx54gGuuuYbzzjuPq666igcffJBbb72Vv/u7v5v223PPPZdf/vKX7LLLLk/DyFdfbLzxxvzyl7/k2muvfbqHkpGRkZGRkbEMMomdkZGRkZHxJFAUBR/72MdYY401APjmN7/5NI/oyeOHP/whd955J2uttVYiLM877zz6/f7TPLKMlUFMh1/VEd4HHngg66+/Pueffz5TU1PTvr/44otZtGgRRx11FMaYGc+xzjrrsO2227LOOuus0rFlLB9rr702H/vYxwB45JFHuOmmm57mET0+ROJ5JhL7Rz/6EZOTk+y///7svPPO3HjjjQwGg6FjrLV8//vfHzrXXwquuOIKli5dykYbbcTHP/5xer3etGN23HFHPvKRj0z7fNNNN2XbbbdldHT0zzHUZwzKsmTbbbdlyy23fLqHkpGRkZGRkbEMMomdkZGRkZHxJNHr9Xj2s58NSGTcsvjOd77DCSecwPbbb88666yTamkffvjhy43cbteOffjhh/mHf/gH/uZv/oZOp8Pf/M3fcMIJJ/DYY48td0znnnsuO++8M6Ojo8ydO5f9998/ETl/CmeccQYARx55JPvuuy9bbbUVCxYs4NJLL53x+JWpIbqiFO57772XN7/5zWywwQbpWX7gAx9gampqufVQ25/ffPPNHHjggay99trMnj2befPmDd3rt771LfbZZx/WWmstZs2axb777svtt9++3LEuXLiQD3zgA2y//fbMnj2b0dFRnve85/HBD36QiYmJacc/kXe11157JULthhtuGKpjvLrWYi2Kgje96U0sXLhwxrkQHR5HH330cs8xU03kBQsWsNlmm6GU4vTTT5/2m6VLl7LtttuilOL//J//8+RvpIX4zAG++MUvsuOOOzI2Nsaaa67JAQccwM0337xKr/d04fnPf376+0wyqq5rzjjjDPbaay/mzp1Lt9tliy224O1vfzv333//jOe85JJLOPbYY3nuc5/LWmutRa/XY4sttuDoo4/mrrvuWmVjj+vkBz/4AXVdD30X5cJee+3FvHnzmJyc5Ec/+tHQMbfddhtLliyh1+ux6667Dn337W9/m4MOOoj11luPTqfDRhttxOGHH86tt94641jacuf73/8+Bx98MOuuuy5aa84+++yVup8rrriCefPmMXv2bObMmcMee+zBN77xjZX67bKI73Ldddd93L9dnmyNNfDPPvts7rrrLg4//HDWW289xsbG2HnnnYfG+qMf/YhDDjmEddddl5GREXbdddcZI5if7B4xEx5++GE+85nPcMABB7DFFlswMjLCGmuswU477cT/+T//Z5qjLdYqv/feewHYYosthuRufA5/aqwPPPAAJ5xwAs9+9rPp9XrMmTOH3XffnS984Qsz1m1v10gfHx/nX/7lX9hqq63odrtssMEGzJ8/n//93/9dqXvOyMjIyMj4a0cmsTMyMjIyMlYBFi9eDMD6668/7bvjjjuO//qv/0Jrze67785BBx3EnDlzuPDCC9ltt91mbLoVcf/99/PCF76Qr3/96+yyyy7su+++LFmyhM997nPst99+qW51G+985zuZP38+t99+OzvvvDMvf/nLuf/++9lrr7247LLLVngfS5Ys4aKLLgKEjFRK8Za3vAVYfhmJJ4Nf/OIX7LTTTpxzzjkYY3jlK1/JNttswyc+8Qn23XffGe+vjSuvvJI99tiDhx56KBHu3/ve99h333258cYb+fznP8+BBx7I1NQU++23HxtvvDHf+c53mDdvHr/5zW9mHM8LXvAC/uM//oM//vGPvOQlL+FlL3sZDz/8MP/2b//G7rvvzqJFi2Ycy+N5V/vvvz8vf/nLAZkzsYbx/Pnzec1rXvMknuhTi0hQLzsXfvvb33LDDTew++67s/XWWz+uc86dO5cLL7yQsiz5//6//4+f/OQnQ9+/9a1v5a677uLAAw/kn/7pn57U+JeHE088kbe97W2Mjo7yyle+kr/5m7/h6quvZo899liu8+aZhCifYLqMWrJkCfvuuy9/93d/x2233cbzn/98DjnkELrdLqeffjo77LDDjPWoX/e613H++eczMjLC3nvvzctf/nK01nzpS19ixx135MYbb1wlY3/xi19Mr9djyZIl0+qxX3/99RhjeMlLXsK8efOA6RHb8d+77ror3W43ff5v//Zv7L///lx11VVsvfXWvOY1r2H99dfnwgsv5MUvfvEK5d1FF13EXnvtxe9+9zte9rKXse+++w6de3n41Kc+xcEHH8z3vvc9tttuuySbDj30UD772c+u9DOJ2HTTTQH42c9+tsrLX9x+++3suOOO3HHHHeyzzz684AUv4NZbb+VVr3oVF198MZdddhl77LEHDzzwAPvssw/bbLMNN998M/vvv/9yG2yuSnz729/mne98Jz/96U/ZbLPNOPTQQ9lll1246667+Od//mf23nvvoeyhrbbaivnz5zM2NgYM146fP38+G2y
"text/plain": [
"<Figure size 1600x1000 with 12 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
2026-06-12 13:49:18 +00:00
"source": [
"vis_ds = tf.keras.utils.image_dataset_from_directory(\n",
" train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)\n",
"\n",
"plt.figure(figsize=(16, 10))\n",
"for images, labels in vis_ds.take(1):\n",
" # Original (4)\n",
" for i in range(4):\n",
" plt.subplot(3, 4, i + 1)\n",
" plt.imshow(images[i].numpy().astype(\"uint8\"))\n",
" plt.title(f\"Asli: {class_names[labels[i].numpy()]}\", fontsize=11)\n",
" plt.axis(\"off\")\n",
" # RandAugment (4)\n",
" aug = _randaug_select(tf.cast(images, tf.float32), ops_per_image=3, magnitude=0.7)\n",
" for i in range(4):\n",
" plt.subplot(3, 4, i + 5)\n",
" plt.imshow(tf.clip_by_value(aug[i], 0, 255).numpy().astype(\"uint8\"))\n",
" plt.title(f\"RandAug: {class_names[labels[i].numpy()]}\", fontsize=11)\n",
" plt.axis(\"off\")\n",
" # MixUp result (4)\n",
" aug_f = tf.cast(images, tf.float32)\n",
" mixed, _ = mix_up(aug_f, tf.one_hot(labels, NUM_CLASSES))\n",
" for i in range(4):\n",
" plt.subplot(3, 4, i + 9)\n",
" plt.imshow(tf.clip_by_value(mixed[i], 0, 255).numpy().astype(\"uint8\"))\n",
" plt.title(\"MixUp/Weather\", fontsize=11)\n",
" plt.axis(\"off\")\n",
"\n",
"plt.suptitle(\"RandAugment + MixUp — Real-World Simulation\", fontsize=16)\n",
"plt.tight_layout()\n",
"plt.show()\n"
]
},
{
"cell_type": "markdown",
"id": "05281076",
"metadata": {},
"source": [
"## 7. Arsitektur Model — CBAM + Lightweight Head\n",
"\n",
"**EfficientNetV2B0** base (frozen awal) + **CBAM spatial attention** + lightweight classifier head.\n",
"\n",
"### Head (≈700K params vs 7.3M sebelumnya)\n",
"```\n",
"Base (7× ×
"```\n",
"\n",
"CBAM (Convolutional Block Attention Module): channel attention + spatial attention\n",
"→ model belajar fokus ke foreground (daun) bukan background.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "4937fb17",
"metadata": {},
"outputs": [],
"source": [
"def cbam_block(x, ratio=8, name=\"cbam\"):\n",
" # Convolutional Block Attention Module — ringan, fokus ke foreground.\n",
" channels = x.shape[-1]\n",
"\n",
" # Channel Attention\n",
" avg_pool = layers.GlobalAveragePooling2D()(x)\n",
" max_pool = layers.GlobalMaxPooling2D()(x)\n",
" ca = layers.Dense(channels // ratio, activation='swish', name=f\"{name}_ca1\")(avg_pool)\n",
" ca = layers.Dense(channels, activation='sigmoid', name=f\"{name}_ca2\")(ca)\n",
" ca2 = layers.Dense(channels // ratio, activation='swish', name=f\"{name}_ca3\")(max_pool)\n",
" ca2 = layers.Dense(channels, activation='sigmoid', name=f\"{name}_ca4\")(ca2)\n",
" ca_out = layers.Add(name=f\"{name}_ca_add\")([ca, ca2])\n",
" ca_out = layers.Reshape((1, 1, channels), name=f\"{name}_ca_reshape\")(ca_out)\n",
" x = layers.Multiply(name=f\"{name}_ca_mul\")([x, ca_out])\n",
"\n",
2026-06-12 15:00:15 +00:00
" # Spatial Attention — output_shape required for Keras 3 serialization\n",
" avg_sp = layers.Lambda(\n",
" lambda t: tf.reduce_mean(t, axis=-1, keepdims=True),\n",
" output_shape=lambda s: s[:-1] + (1,),\n",
" name=f\"{name}_sa_avg\",\n",
" )(x)\n",
" max_sp = layers.Lambda(\n",
" lambda t: tf.reduce_max(t, axis=-1, keepdims=True),\n",
" output_shape=lambda s: s[:-1] + (1,),\n",
" name=f\"{name}_sa_max\",\n",
" )(x)\n",
2026-06-12 13:49:18 +00:00
" sp = layers.Concatenate(name=f\"{name}_sa_cat\")([avg_sp, max_sp])\n",
" sp = layers.Conv2D(1, 7, padding='same', activation='sigmoid', name=f\"{name}_sa_conv\")(sp)\n",
" x = layers.Multiply(name=f\"{name}_sa_mul\")([x, sp])\n",
" return x\n",
"\n",
"def build_model(num_classes, target_size=(224, 224)):\n",
" # Accept any input size via variable input; Resizing handles progressive resolution.\n",
" inputs = tf.keras.Input(shape=(None, None, 3), name=\"input\")\n",
" x = layers.Resizing(target_size[0], target_size[1], interpolation='bilinear',\n",
" name=\"resize_input\")(inputs)\n",
"\n",
" base_model = EfficientNetV2B0(\n",
" input_shape=target_size + (3,),\n",
" include_top=False,\n",
" weights='imagenet',\n",
" )\n",
" base_model.trainable = False\n",
"\n",
" # Gaussian noise untuk regularisasi\n",
" x = layers.GaussianNoise(0.05, name=\"gauss_noise\")(x)\n",
" x = base_model(x, training=False)\n",
" # CBAM attention — fokus ke region daun\n",
" x = cbam_block(x, ratio=8, name=\"cbam\")\n",
" x = layers.GlobalAveragePooling2D(name=\"gap\")(x)\n",
" x = layers.Dropout(0.3, name=\"drop_gap\")(x)\n",
" x = layers.Dense(512, activation='swish', name=\"dense_head\")(x)\n",
" x = layers.BatchNormalization(name=\"bn_head\")(x)\n",
" x = layers.Dropout(0.4, name=\"drop_head\")(x)\n",
" outputs = layers.Dense(num_classes, activation='linear', dtype='float32', name=\"logits\")(x)\n",
" return models.Model(inputs, outputs), base_model\n",
"\n",
"# Checkpoint\n",
"ckpt_dir = '/content/best_model' if IS_COLAB else os.path.join(os.getcwd(), 'best_model')\n",
"checkpoint_path = os.path.join(ckpt_dir, 'best_model.keras')\n",
"\n",
"# Hapus checkpoint lama (arsitektur berbeda — tidak kompatibel)\n",
"if os.path.exists(checkpoint_path):\n",
" print(f\"Removing old checkpoint (incompatible architecture)...\")\n",
" os.remove(checkpoint_path)\n",
"\n",
"if os.path.exists(checkpoint_path):\n",
" print(f\"Loading checkpoint: {checkpoint_path}\")\n",
" try:\n",
" model = models.load_model(checkpoint_path, compile=False)\n",
" except Exception as e:\n",
" print(f\"Load failed: {e}. Building fresh.\")\n",
" model, base_model = build_model(NUM_CLASSES)\n",
"else:\n",
" print(\"No checkpoint. Building fresh model.\")\n",
" model, base_model = build_model(NUM_CLASSES)\n",
"\n",
"os.makedirs(ckpt_dir, exist_ok=True)\n",
2026-06-12 15:00:15 +00:00
"model.summary()"
2026-06-12 13:49:18 +00:00
]
},
{
"cell_type": "markdown",
"id": "28aaaed1",
"metadata": {},
"source": [
"## 8. Training Setup — Cosine Decay + SWA + Callbacks\n",
"\n",
"Mengganti `ReduceLROnPlateau` / `EarlyStopping` dengan:\n",
"- **CosineDecay** + linear warmup per fase\n",
"- **Stochastic Weight Averaging (SWA)** — averaging bobot untuk wider optima\n",
"- **ModelCheckpoint** — tetap simpan best val_accuracy\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 9,
2026-06-12 13:49:18 +00:00
"id": "51deff9d",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Callbacks ready.\n",
"Checkpoint path: /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"class WarmupCosineDecay(tf.keras.optimizers.schedules.LearningRateSchedule):\n",
" # Cosine decay with linear warmup.\n",
" def __init__(self, warmup_steps, total_steps, peak_lr, min_lr=1e-7):\n",
" super().__init__()\n",
" self.warmup_steps = warmup_steps\n",
" self.total_steps = total_steps\n",
" self.peak_lr = peak_lr\n",
" self.min_lr = min_lr\n",
"\n",
" def __call__(self, step):\n",
" step = tf.cast(step, tf.float32)\n",
" warmup_steps = tf.cast(self.warmup_steps, tf.float32)\n",
" total_steps = tf.cast(self.total_steps, tf.float32)\n",
" # Warmup phase\n",
" warmup_lr = self.peak_lr * (step / warmup_steps)\n",
" # Cosine decay phase\n",
" progress = (step - warmup_steps) / tf.maximum(total_steps - warmup_steps, 1.0)\n",
" cosine_lr = self.min_lr + 0.5 * (self.peak_lr - self.min_lr) * (1.0 + tf.cos(np.pi * progress))\n",
" return tf.where(step < warmup_steps, warmup_lr, cosine_lr)\n",
"\n",
" def get_config(self):\n",
" return {\n",
" \"warmup_steps\": self.warmup_steps, \"total_steps\": self.total_steps,\n",
" \"peak_lr\": self.peak_lr, \"min_lr\": self.min_lr,\n",
" }\n",
"\n",
"class SWACallback(tf.keras.callbacks.Callback):\n",
" # Stochastic Weight Averaging — averages weights over final epochs.\n",
" def __init__(self, start_epoch, swa_lr=1e-5):\n",
" super().__init__()\n",
" self.start_epoch = start_epoch\n",
" self.swa_lr = swa_lr\n",
" self.swa_weights = None\n",
" self.n_models = 0\n",
"\n",
" def on_epoch_begin(self, epoch, logs=None):\n",
" if epoch >= self.start_epoch and self.swa_weights is None:\n",
" self.swa_weights = [w.numpy() for w in self.model.weights]\n",
" print(f\"\\nSWA: starting weight averaging at epoch {epoch+1}\")\n",
"\n",
" def on_epoch_end(self, epoch, logs=None):\n",
" if epoch >= self.start_epoch and self.swa_weights is not None:\n",
" for i, w in enumerate(self.model.weights):\n",
" self.swa_weights[i] = (self.swa_weights[i] * self.n_models + w.numpy()) / (self.n_models + 1)\n",
" self.n_models += 1\n",
"\n",
" def apply_swa_weights(self):\n",
" if self.swa_weights is None:\n",
" print(\"SWA: no weights to average (skipped)\")\n",
" return\n",
" for w, swa_w in zip(self.model.weights, self.swa_weights):\n",
" w.assign(swa_w)\n",
" print(f\"SWA weights applied ({self.n_models} models averaged).\")\n",
"\n",
"# Shared callbacks\n",
"checkpoint_cb = callbacks.ModelCheckpoint(\n",
" checkpoint_path, save_best_only=True, monitor=\"val_accuracy\",\n",
" mode=\"max\", verbose=1)\n",
"\n",
"csv_logger = callbacks.CSVLogger(os.path.join(ckpt_dir, 'training_log.csv'))\n",
"\n",
"def make_callbacks(swa_start=None):\n",
" cbs = [checkpoint_cb, csv_logger]\n",
" if swa_start is not None:\n",
" cbs.append(SWACallback(swa_start))\n",
" return cbs\n",
"\n",
"print(\"Callbacks ready.\")\n",
"print(f\"Checkpoint path: {checkpoint_path}\")\n"
]
},
{
"cell_type": "markdown",
"id": "82814b7b",
"metadata": {},
"source": [
"## 9. Fase 1 — Head Only Training (128×
"\n",
"Progressive resolution: mulai dari **128×
"Base model beku, hanya head (CBAM + Dense) yang dilatih.\n",
"Optimizer: AdamW + EMA + CosineDecay(warmup=3, peak=1e-3).\n",
"Label smoothing: 0.15\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 10,
2026-06-12 13:49:18 +00:00
"id": "decaf5cb",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 3526 files belonging to 4 classes.\n",
"Found 754 files belonging to 4 classes.\n",
"Phase 1: Head training at 128×
"Epoch 1/25\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
"I0000 00:00:1781273809.189453 1199382 service.cc:152] XLA service 0x757cd4001ab0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
"I0000 00:00:1781273809.189482 1199382 service.cc:160] StreamExecutor device (0): NVIDIA A100-SXM4-40GB, Compute Capability 8.0\n",
"2026-06-12 14:16:50.547970: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
"I0000 00:00:1781273813.442292 1199382 cuda_dnn.cc:529] Loaded cuDNN version 92301\n",
"2026-06-12 14:16:56.177029: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 12 bytes spill stores, 12 bytes spill loads\n",
"\n",
"2026-06-12 14:16:56.187342: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 232 bytes spill stores, 232 bytes spill loads\n",
"\n",
"2026-06-12 14:16:56.224294: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080_0', 428 bytes spill stores, 1360 bytes spill loads\n",
"\n",
"2026-06-12 14:16:56.985744: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 5372 bytes spill stores, 5348 bytes spill loads\n",
"\n",
"2026-06-12 14:16:57.807703: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 6000 bytes spill stores, 5952 bytes spill loads\n",
"\n",
"2026-06-12 14:17:00.117091: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:00.273419: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:01.074555: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:01.238208: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:01.856060: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:02.007414: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:02.380303: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:02.547537: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:03.217188: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:03.384704: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:03.827983: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:03.984354: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:04.179865: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:04.337508: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m 11/111\u001b[0m \u001b[32m━\u001b[0m\u001b[37m━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[1m1s\u001b[0m 12ms/step - accuracy: 0.2155 - loss: 2.4408 "
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"I0000 00:00:1781273833.778844 1199382 device_compiler.h:188] Compiled cluster using XLA! This line is logged at most once for the lifetime of the process.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 46ms/step - accuracy: 0.4442 - loss: 1.7988"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:17:23.251739: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 156 bytes spill stores, 156 bytes spill loads\n",
"\n",
"2026-06-12 14:17:23.700939: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 5100 bytes spill stores, 5108 bytes spill loads\n",
"\n",
"2026-06-12 14:17:24.225262: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_9080', 4500 bytes spill stores, 4492 bytes spill loads\n",
"\n",
"2026-06-12 14:17:25.737787: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:25.891964: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:26.674550: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:26.835098: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:27.310738: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:27.461542: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 209ms/step - accuracy: 0.4475 - loss: 1.7909"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:17:42.009559: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 232 bytes spill stores, 232 bytes spill loads\n",
"\n",
"2026-06-12 14:17:42.164736: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591_0', 496 bytes spill stores, 1400 bytes spill loads\n",
"\n",
"2026-06-12 14:17:42.197096: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 12 bytes spill stores, 12 bytes spill loads\n",
"\n",
"2026-06-12 14:17:42.810303: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 6000 bytes spill stores, 5952 bytes spill loads\n",
"\n",
"2026-06-12 14:17:43.028755: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 5372 bytes spill stores, 5348 bytes spill loads\n",
"\n",
"2026-06-12 14:17:47.498104: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2484', 252 bytes spill stores, 252 bytes spill loads\n",
"\n",
"2026-06-12 14:17:47.518030: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2484', 16 bytes spill stores, 16 bytes spill loads\n",
"\n",
"2026-06-12 14:17:47.830502: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 252 bytes spill stores, 252 bytes spill loads\n",
"\n",
"2026-06-12 14:17:47.831081: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591_0', 492 bytes spill stores, 1464 bytes spill loads\n",
"\n",
"2026-06-12 14:17:47.902990: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 16 bytes spill stores, 16 bytes spill loads\n",
"\n",
"2026-06-12 14:17:48.086926: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2484', 5912 bytes spill stores, 5980 bytes spill loads\n",
"\n",
"2026-06-12 14:17:48.392613: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2484', 5356 bytes spill stores, 5344 bytes spill loads\n",
"\n",
"2026-06-12 14:17:48.398217: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 5356 bytes spill stores, 5344 bytes spill loads\n",
"\n",
"2026-06-12 14:17:48.749904: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2591', 5912 bytes spill stores, 5980 bytes spill loads\n",
"\n",
"2026-06-12 14:17:50.146437: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:50.302796: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:51.093951: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:51.257390: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:51.861787: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:52.013103: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:52.381425: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:52.548282: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:53.461750: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:53.617936: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:53.814377: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:17:53.972623: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"Epoch 1: val_accuracy improved from None to 0.97215, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 1: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m80s\u001b[0m 400ms/step - accuracy: 0.6322 - loss: 1.3592 - val_accuracy: 0.9721 - val_loss: 0.8272\n",
"Epoch 2/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 70ms/step - accuracy: 0.8167 - loss: 0.9964\n",
"Epoch 2: val_accuracy improved from 0.97215 to 0.98276, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 2: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 100ms/step - accuracy: 0.8267 - loss: 0.9727 - val_accuracy: 0.9828 - val_loss: 0.7051\n",
"Epoch 3/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 73ms/step - accuracy: 0.8284 - loss: 0.9292\n",
"Epoch 3: val_accuracy did not improve from 0.98276\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 96ms/step - accuracy: 0.8542 - loss: 0.8379 - val_accuracy: 0.7599 - val_loss: 1.0449\n",
"Epoch 4/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 74ms/step - accuracy: 0.8979 - loss: 0.7535\n",
"Epoch 4: val_accuracy did not improve from 0.98276\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 95ms/step - accuracy: 0.9058 - loss: 0.7323 - val_accuracy: 0.9668 - val_loss: 0.8505\n",
"Epoch 5/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 75ms/step - accuracy: 0.9256 - loss: 0.7010\n",
"Epoch 5: val_accuracy did not improve from 0.98276\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 98ms/step - accuracy: 0.9280 - loss: 0.6952 - val_accuracy: 0.9814 - val_loss: 0.6596\n",
"Epoch 6/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 72ms/step - accuracy: 0.9273 - loss: 0.6984\n",
"Epoch 6: val_accuracy improved from 0.98276 to 0.98806, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 6: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 105ms/step - accuracy: 0.9348 - loss: 0.6864 - val_accuracy: 0.9881 - val_loss: 0.5717\n",
"Epoch 7/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 75ms/step - accuracy: 0.9340 - loss: 0.6884\n",
"Epoch 7: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 98ms/step - accuracy: 0.9317 - loss: 0.6804 - val_accuracy: 0.9867 - val_loss: 0.5296\n",
"Epoch 8/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 69ms/step - accuracy: 0.9318 - loss: 0.6772\n",
"Epoch 8: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 92ms/step - accuracy: 0.9305 - loss: 0.6687 - val_accuracy: 0.9854 - val_loss: 0.5359\n",
"Epoch 9/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 74ms/step - accuracy: 0.9356 - loss: 0.6738\n",
"Epoch 9: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 97ms/step - accuracy: 0.9393 - loss: 0.6683 - val_accuracy: 0.9841 - val_loss: 0.5261\n",
"Epoch 10/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 73ms/step - accuracy: 0.9334 - loss: 0.6756\n",
"Epoch 10: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 96ms/step - accuracy: 0.9348 - loss: 0.6637 - val_accuracy: 0.9867 - val_loss: 0.5263\n",
"Epoch 11/25\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 67ms/step - accuracy: 0.9426 - loss: 0.6642\n",
"Epoch 11: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 90ms/step - accuracy: 0.9433 - loss: 0.6501 - val_accuracy: 0.9854 - val_loss: 0.5230\n",
"Epoch 12/25\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 66ms/step - accuracy: 0.9521 - loss: 0.6532\n",
"Epoch 12: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 91ms/step - accuracy: 0.9472 - loss: 0.6467 - val_accuracy: 0.9867 - val_loss: 0.5194\n",
"Epoch 13/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 66ms/step - accuracy: 0.9477 - loss: 0.6544\n",
"Epoch 13: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 89ms/step - accuracy: 0.9475 - loss: 0.6505 - val_accuracy: 0.9841 - val_loss: 0.5208\n",
"Epoch 14/25\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 76ms/step - accuracy: 0.9464 - loss: 0.6560\n",
"Epoch 14: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 97ms/step - accuracy: 0.9444 - loss: 0.6519 - val_accuracy: 0.9854 - val_loss: 0.5189\n",
"Epoch 15/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 69ms/step - accuracy: 0.9446 - loss: 0.6534\n",
"Epoch 15: val_accuracy did not improve from 0.98806\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 91ms/step - accuracy: 0.9467 - loss: 0.6435 - val_accuracy: 0.9828 - val_loss: 0.5195\n",
"Epoch 16/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 69ms/step - accuracy: 0.9448 - loss: 0.6475\n",
"Epoch 16: val_accuracy improved from 0.98806 to 0.98939, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 16: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 97ms/step - accuracy: 0.9433 - loss: 0.6397 - val_accuracy: 0.9894 - val_loss: 0.5242\n",
"Epoch 17/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 70ms/step - accuracy: 0.9541 - loss: 0.6491\n",
"Epoch 17: val_accuracy improved from 0.98939 to 0.99072, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 17: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 100ms/step - accuracy: 0.9535 - loss: 0.6364 - val_accuracy: 0.9907 - val_loss: 0.5150\n",
"Epoch 18/25\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 75ms/step - accuracy: 0.9404 - loss: 0.6535\n",
"Epoch 18: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 96ms/step - accuracy: 0.9427 - loss: 0.6423 - val_accuracy: 0.9894 - val_loss: 0.5153\n",
"Epoch 19/25\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 66ms/step - accuracy: 0.9530 - loss: 0.6359\n",
"Epoch 19: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 88ms/step - accuracy: 0.9492 - loss: 0.6301 - val_accuracy: 0.9881 - val_loss: 0.5137\n",
"Epoch 20/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 62ms/step - accuracy: 0.9509 - loss: 0.6385\n",
"Epoch 20: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 83ms/step - accuracy: 0.9535 - loss: 0.6307 - val_accuracy: 0.9881 - val_loss: 0.5132\n",
"Epoch 21/25\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 62ms/step - accuracy: 0.9476 - loss: 0.6314\n",
"Epoch 21: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 87ms/step - accuracy: 0.9512 - loss: 0.6277 - val_accuracy: 0.9894 - val_loss: 0.5152\n",
"Epoch 22/25\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 67ms/step - accuracy: 0.9507 - loss: 0.6502\n",
"Epoch 22: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 92ms/step - accuracy: 0.9538 - loss: 0.6294 - val_accuracy: 0.9867 - val_loss: 0.5134\n",
"Epoch 23/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 66ms/step - accuracy: 0.9501 - loss: 0.6412\n",
"Epoch 23: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 88ms/step - accuracy: 0.9509 - loss: 0.6314 - val_accuracy: 0.9867 - val_loss: 0.5125\n",
"Epoch 24/25\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 78ms/step - accuracy: 0.9551 - loss: 0.6356\n",
"Epoch 24: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 99ms/step - accuracy: 0.9546 - loss: 0.6216 - val_accuracy: 0.9867 - val_loss: 0.5123\n",
"Epoch 25/25\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 68ms/step - accuracy: 0.9517 - loss: 0.6328\n",
"Epoch 25: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 93ms/step - accuracy: 0.9546 - loss: 0.6220 - val_accuracy: 0.9867 - val_loss: 0.5126\n"
]
}
],
"source": [
"IMG_128 = (128, 128)\n",
"\n",
"# Rebuild datasets at 128x128\n",
"train_ds_128 = tf.keras.utils.image_dataset_from_directory(\n",
" train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_128)\n",
"val_ds_128 = tf.keras.utils.image_dataset_from_directory(\n",
" val_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_128)\n",
"\n",
"train_ds_128 = (train_ds_128.map(augment_and_mix, num_parallel_calls=AUTOTUNE)\n",
" .map(preprocess_fn, num_parallel_calls=AUTOTUNE)\n",
" .map(add_sample_weight, num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"val_ds_128 = (val_ds_128.map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)\n",
" .map(lambda img, lbl: (img, tf.one_hot(lbl, NUM_CLASSES)), num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"\n",
"EPOCHS_P1 = 25\n",
"card = tf.data.experimental.cardinality(train_ds_128).numpy()\n",
"steps_per_epoch = card if card > 0 else 100\n",
"total_steps = steps_per_epoch * EPOCHS_P1\n",
"warmup_steps = steps_per_epoch * 3 # 3 epoch warmup\n",
"\n",
"lr_schedule_p1 = WarmupCosineDecay(warmup_steps, total_steps, peak_lr=1e-3, min_lr=1e-5)\n",
"\n",
"model.compile(\n",
" optimizer=AdamW(\n",
" learning_rate=lr_schedule_p1, weight_decay=1e-4),\n",
" loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True, label_smoothing=0.15),\n",
" metrics=['accuracy']\n",
")\n",
"\n",
"print(\"Phase 1: Head training at 128×
"history_1 = model.fit(train_ds_128, validation_data=val_ds_128,\n",
" epochs=EPOCHS_P1, callbacks=make_callbacks())"
]
2026-06-12 13:49:18 +00:00
},
{
"cell_type": "markdown",
"id": "de673df6",
"metadata": {},
"source": [
"## 10. Fase 2 — Partial Fine-tuning (192×
"\n",
"Resolusi naik ke **192×
"Learning rate lebih rendah: peak=5e-4, cosine decay ke 1e-6.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 11,
2026-06-12 13:49:18 +00:00
"id": "fd7fb926",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 3526 files belonging to 4 classes.\n",
"Found 754 files belonging to 4 classes.\n",
"Phase 2: Fine-tuning top 100 layers at 192×
"Epoch 26/55\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:23:00.868983: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_20707', 380 bytes spill stores, 380 bytes spill loads\n",
"\n",
"2026-06-12 14:23:00.892104: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_20707', 532 bytes spill stores, 532 bytes spill loads\n",
"\n",
"2026-06-12 14:23:02.414776: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:23:02.572580: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 37ms/step - accuracy: 0.8958 - loss: 0.8296"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:23:26.656239: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_20707', 284 bytes spill stores, 284 bytes spill loads\n",
"\n",
"2026-06-12 14:23:26.707546: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_20707', 204 bytes spill stores, 204 bytes spill loads\n",
"\n",
"2026-06-12 14:23:28.284698: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:23:28.442489: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 227ms/step - accuracy: 0.8959 - loss: 0.8289\n",
"Epoch 26: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m83s\u001b[0m 324ms/step - accuracy: 0.9141 - loss: 0.7556 - val_accuracy: 0.9854 - val_loss: 0.5342\n",
"Epoch 27/55\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 49ms/step - accuracy: 0.9321 - loss: 0.6812\n",
"Epoch 27: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9353 - loss: 0.6634 - val_accuracy: 0.9867 - val_loss: 0.5301\n",
"Epoch 28/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 50ms/step - accuracy: 0.9360 - loss: 0.6741\n",
"Epoch 28: val_accuracy did not improve from 0.99072\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9419 - loss: 0.6621 - val_accuracy: 0.9828 - val_loss: 0.5267\n",
"Epoch 29/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 49ms/step - accuracy: 0.9351 - loss: 0.6449\n",
"Epoch 29: val_accuracy improved from 0.99072 to 0.99469, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 29: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 75ms/step - accuracy: 0.9433 - loss: 0.6353 - val_accuracy: 0.9947 - val_loss: 0.5227\n",
"Epoch 30/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 48ms/step - accuracy: 0.9405 - loss: 0.6331\n",
"Epoch 30: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 65ms/step - accuracy: 0.9498 - loss: 0.6274 - val_accuracy: 0.9894 - val_loss: 0.5211\n",
"Epoch 31/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 43ms/step - accuracy: 0.9568 - loss: 0.6301\n",
"Epoch 31: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 59ms/step - accuracy: 0.9563 - loss: 0.6217 - val_accuracy: 0.9947 - val_loss: 0.5088\n",
"Epoch 32/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 44ms/step - accuracy: 0.9485 - loss: 0.6269\n",
"Epoch 32: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 62ms/step - accuracy: 0.9532 - loss: 0.6204 - val_accuracy: 0.9920 - val_loss: 0.5113\n",
"Epoch 33/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9529 - loss: 0.6155\n",
"Epoch 33: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 66ms/step - accuracy: 0.9566 - loss: 0.6145 - val_accuracy: 0.9920 - val_loss: 0.5079\n",
"Epoch 34/55\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 42ms/step - accuracy: 0.9579 - loss: 0.6217\n",
"Epoch 34: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 58ms/step - accuracy: 0.9609 - loss: 0.6066 - val_accuracy: 0.9947 - val_loss: 0.5062\n",
"Epoch 35/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 46ms/step - accuracy: 0.9488 - loss: 0.6275\n",
"Epoch 35: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 62ms/step - accuracy: 0.9569 - loss: 0.6065 - val_accuracy: 0.9947 - val_loss: 0.5055\n",
"Epoch 36/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 46ms/step - accuracy: 0.9513 - loss: 0.6225\n",
"Epoch 36: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 64ms/step - accuracy: 0.9589 - loss: 0.6180 - val_accuracy: 0.9934 - val_loss: 0.5105\n",
"Epoch 37/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9574 - loss: 0.6120\n",
"Epoch 37: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 67ms/step - accuracy: 0.9626 - loss: 0.6054 - val_accuracy: 0.9920 - val_loss: 0.5078\n",
"Epoch 38/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 41ms/step - accuracy: 0.9620 - loss: 0.6133\n",
"Epoch 38: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 56ms/step - accuracy: 0.9637 - loss: 0.6065 - val_accuracy: 0.9934 - val_loss: 0.5003\n",
"Epoch 39/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 44ms/step - accuracy: 0.9657 - loss: 0.6051\n",
"Epoch 39: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 62ms/step - accuracy: 0.9657 - loss: 0.5989 - val_accuracy: 0.9920 - val_loss: 0.5097\n",
"Epoch 40/55\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 42ms/step - accuracy: 0.9592 - loss: 0.5924\n",
"Epoch 40: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 60ms/step - accuracy: 0.9631 - loss: 0.5926 - val_accuracy: 0.9934 - val_loss: 0.5028\n",
"Epoch 41/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 44ms/step - accuracy: 0.9630 - loss: 0.6003\n",
"Epoch 41: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 59ms/step - accuracy: 0.9682 - loss: 0.5915 - val_accuracy: 0.9947 - val_loss: 0.4992\n",
"Epoch 42/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 44ms/step - accuracy: 0.9668 - loss: 0.5889\n",
"Epoch 42: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 59ms/step - accuracy: 0.9688 - loss: 0.5840 - val_accuracy: 0.9947 - val_loss: 0.4995\n",
"Epoch 43/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 46ms/step - accuracy: 0.9640 - loss: 0.6092\n",
"Epoch 43: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 61ms/step - accuracy: 0.9640 - loss: 0.5933 - val_accuracy: 0.9947 - val_loss: 0.4971\n",
"Epoch 44/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 48ms/step - accuracy: 0.9562 - loss: 0.6077\n",
"Epoch 44: val_accuracy did not improve from 0.99469\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 66ms/step - accuracy: 0.9594 - loss: 0.6028 - val_accuracy: 0.9947 - val_loss: 0.5002\n",
"Epoch 45/55\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 43ms/step - accuracy: 0.9620 - loss: 0.5919\n",
"Epoch 45: val_accuracy improved from 0.99469 to 0.99602, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\n",
"Epoch 45: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 67ms/step - accuracy: 0.9697 - loss: 0.5841 - val_accuracy: 0.9960 - val_loss: 0.4982\n",
"Epoch 46/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 47ms/step - accuracy: 0.9607 - loss: 0.5942\n",
"Epoch 46: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 62ms/step - accuracy: 0.9702 - loss: 0.5871 - val_accuracy: 0.9960 - val_loss: 0.4972\n",
"Epoch 47/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 48ms/step - accuracy: 0.9634 - loss: 0.5912\n",
"Epoch 47: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 64ms/step - accuracy: 0.9680 - loss: 0.5859 - val_accuracy: 0.9960 - val_loss: 0.4970\n",
"Epoch 48/55\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 47ms/step - accuracy: 0.9643 - loss: 0.5886\n",
"Epoch 48: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 61ms/step - accuracy: 0.9680 - loss: 0.5828 - val_accuracy: 0.9947 - val_loss: 0.4962\n",
"Epoch 49/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 45ms/step - accuracy: 0.9689 - loss: 0.5773\n",
"Epoch 49: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 63ms/step - accuracy: 0.9733 - loss: 0.5714 - val_accuracy: 0.9947 - val_loss: 0.4984\n",
"Epoch 50/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 47ms/step - accuracy: 0.9692 - loss: 0.5851\n",
"Epoch 50: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 62ms/step - accuracy: 0.9722 - loss: 0.5835 - val_accuracy: 0.9947 - val_loss: 0.4955\n",
"Epoch 51/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 43ms/step - accuracy: 0.9720 - loss: 0.5699\n",
"Epoch 51: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 58ms/step - accuracy: 0.9716 - loss: 0.5712 - val_accuracy: 0.9960 - val_loss: 0.4965\n",
"Epoch 52/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9703 - loss: 0.5732\n",
"Epoch 52: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 69ms/step - accuracy: 0.9688 - loss: 0.5746 - val_accuracy: 0.9947 - val_loss: 0.4965\n",
"Epoch 53/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 42ms/step - accuracy: 0.9713 - loss: 0.5905\n",
"Epoch 53: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 58ms/step - accuracy: 0.9711 - loss: 0.5873 - val_accuracy: 0.9947 - val_loss: 0.4969\n",
"Epoch 54/55\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 45ms/step - accuracy: 0.9729 - loss: 0.5757\n",
"Epoch 54: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 60ms/step - accuracy: 0.9736 - loss: 0.5732 - val_accuracy: 0.9960 - val_loss: 0.4953\n",
"Epoch 55/55\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 44ms/step - accuracy: 0.9723 - loss: 0.5850\n",
"Epoch 55: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 60ms/step - accuracy: 0.9759 - loss: 0.5766 - val_accuracy: 0.9947 - val_loss: 0.4956\n"
]
}
],
"source": [
"IMG_192 = (192, 192)\n",
"\n",
"train_ds_192 = tf.keras.utils.image_dataset_from_directory(\n",
" train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_192)\n",
"val_ds_192 = tf.keras.utils.image_dataset_from_directory(\n",
" val_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_192)\n",
"\n",
"train_ds_192 = (train_ds_192.map(augment_and_mix, num_parallel_calls=AUTOTUNE)\n",
" .map(preprocess_fn, num_parallel_calls=AUTOTUNE)\n",
" .map(add_sample_weight, num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"val_ds_192 = (val_ds_192.map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)\n",
" .map(lambda img, lbl: (img, tf.one_hot(lbl, NUM_CLASSES)), num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"\n",
"# Unfreeze top 100 layers\n",
"base_model.trainable = True\n",
"for layer in base_model.layers[:-100]:\n",
" layer.trainable = False\n",
"\n",
"EPOCHS_P2 = 30\n",
"card_p2 = tf.data.experimental.cardinality(train_ds_192).numpy()\n",
"steps_p2 = card_p2 if card_p2 > 0 else 100\n",
"total_p2 = steps_p2 * EPOCHS_P2\n",
"warmup_p2 = steps_p2 * 2\n",
"\n",
"lr_schedule_p2 = WarmupCosineDecay(warmup_p2, total_p2, peak_lr=5e-4, min_lr=1e-6)\n",
"\n",
"model.compile(\n",
" optimizer=AdamW(\n",
" learning_rate=lr_schedule_p2, weight_decay=1e-4),\n",
" loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True, label_smoothing=0.15),\n",
" metrics=['accuracy']\n",
")\n",
"\n",
"print(\"Phase 2: Fine-tuning top 100 layers at 192×
"history_2 = model.fit(train_ds_192, validation_data=val_ds_192,\n",
" epochs=EPOCHS_P1 + EPOCHS_P2, initial_epoch=history_1.epoch[-1] + 1,\n",
" callbacks=make_callbacks())"
]
2026-06-12 13:49:18 +00:00
},
{
"cell_type": "markdown",
"id": "33326d9c",
"metadata": {},
"source": [
"## 11. Fase 3 — Full Fine-tuning (224×
"\n",
"Resolusi penuh **224×
"LR sangat rendah: peak=1e-4, cosine decay ke 1e-7.\n",
"Label smoothing diturunkan ke 0.10 untuk kalibrasi lebih baik.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 12,
2026-06-12 13:49:18 +00:00
"id": "14115063",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 3526 files belonging to 4 classes.\n",
"Found 754 files belonging to 4 classes.\n",
"Phase 3: Full fine-tuning at 224×
"Epoch 56/85\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:28:37.646510: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:28:37.797973: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:28:38.289799: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:28:38.456772: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:28:42.979933: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:28:43.131707: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 43ms/step - accuracy: 0.9166 - loss: 0.5839"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:29:24.609560: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:29:24.760438: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:29:28.524940: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:29:28.675667: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 414ms/step - accuracy: 0.9166 - loss: 0.5834\n",
"Epoch 56: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m144s\u001b[0m 515ms/step - accuracy: 0.9209 - loss: 0.5566 - val_accuracy: 0.9841 - val_loss: 0.4012\n",
"Epoch 57/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9491 - loss: 0.5064\n",
"Epoch 57: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 67ms/step - accuracy: 0.9552 - loss: 0.5008 - val_accuracy: 0.9881 - val_loss: 0.3947\n",
"Epoch 58/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 59ms/step - accuracy: 0.9571 - loss: 0.5114\n",
"Epoch 58: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 75ms/step - accuracy: 0.9552 - loss: 0.5000 - val_accuracy: 0.9907 - val_loss: 0.3781\n",
"Epoch 59/85\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9583 - loss: 0.5124\n",
"Epoch 59: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 72ms/step - accuracy: 0.9569 - loss: 0.4984 - val_accuracy: 0.9907 - val_loss: 0.3811\n",
"Epoch 60/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 48ms/step - accuracy: 0.9524 - loss: 0.4857\n",
"Epoch 60: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 67ms/step - accuracy: 0.9594 - loss: 0.4783 - val_accuracy: 0.9934 - val_loss: 0.3730\n",
"Epoch 61/85\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 58ms/step - accuracy: 0.9712 - loss: 0.4700\n",
"Epoch 61: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 77ms/step - accuracy: 0.9722 - loss: 0.4640 - val_accuracy: 0.9947 - val_loss: 0.3712\n",
"Epoch 62/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 53ms/step - accuracy: 0.9667 - loss: 0.4793\n",
"Epoch 62: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9654 - loss: 0.4779 - val_accuracy: 0.9934 - val_loss: 0.3734\n",
"Epoch 63/85\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9632 - loss: 0.4863\n",
"Epoch 63: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9657 - loss: 0.4826 - val_accuracy: 0.9947 - val_loss: 0.3709\n",
"Epoch 64/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 53ms/step - accuracy: 0.9634 - loss: 0.4818\n",
"Epoch 64: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9665 - loss: 0.4714 - val_accuracy: 0.9947 - val_loss: 0.3725\n",
"Epoch 65/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 53ms/step - accuracy: 0.9695 - loss: 0.4804\n",
"Epoch 65: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9688 - loss: 0.4724 - val_accuracy: 0.9947 - val_loss: 0.3707\n",
"Epoch 66/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 58ms/step - accuracy: 0.9651 - loss: 0.4806\n",
"Epoch 66: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 74ms/step - accuracy: 0.9697 - loss: 0.4700 - val_accuracy: 0.9934 - val_loss: 0.3715\n",
"Epoch 67/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 55ms/step - accuracy: 0.9713 - loss: 0.4724\n",
"Epoch 67: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9736 - loss: 0.4611 - val_accuracy: 0.9920 - val_loss: 0.3733\n",
"Epoch 68/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 55ms/step - accuracy: 0.9668 - loss: 0.4699\n",
"Epoch 68: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9711 - loss: 0.4644 - val_accuracy: 0.9920 - val_loss: 0.3741\n",
"Epoch 69/85\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 50ms/step - accuracy: 0.9699 - loss: 0.4646\n",
"Epoch 69: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 69ms/step - accuracy: 0.9736 - loss: 0.4569 - val_accuracy: 0.9947 - val_loss: 0.3703\n",
"Epoch 70/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9680 - loss: 0.4670\n",
"Epoch 70: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9748 - loss: 0.4617 - val_accuracy: 0.9947 - val_loss: 0.3716\n",
"Epoch 71/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 56ms/step - accuracy: 0.9709 - loss: 0.4679\n",
"Epoch 71: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9756 - loss: 0.4645 - val_accuracy: 0.9947 - val_loss: 0.3694\n",
"Epoch 72/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 55ms/step - accuracy: 0.9719 - loss: 0.4562\n",
"Epoch 72: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9725 - loss: 0.4569 - val_accuracy: 0.9947 - val_loss: 0.3692\n",
"Epoch 73/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 56ms/step - accuracy: 0.9634 - loss: 0.4767\n",
"Epoch 73: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 74ms/step - accuracy: 0.9716 - loss: 0.4705 - val_accuracy: 0.9934 - val_loss: 0.3697\n",
"Epoch 74/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9743 - loss: 0.4731\n",
"Epoch 74: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 68ms/step - accuracy: 0.9745 - loss: 0.4600 - val_accuracy: 0.9947 - val_loss: 0.3685\n",
"Epoch 75/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9716 - loss: 0.4807\n",
"Epoch 75: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 75ms/step - accuracy: 0.9748 - loss: 0.4649 - val_accuracy: 0.9947 - val_loss: 0.3684\n",
"Epoch 76/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9691 - loss: 0.4631\n",
"Epoch 76: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9702 - loss: 0.4566 - val_accuracy: 0.9947 - val_loss: 0.3695\n",
"Epoch 77/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 53ms/step - accuracy: 0.9679 - loss: 0.4810\n",
"Epoch 77: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9691 - loss: 0.4593 - val_accuracy: 0.9947 - val_loss: 0.3695\n",
"Epoch 78/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 62ms/step - accuracy: 0.9734 - loss: 0.4507\n",
"Epoch 78: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 79ms/step - accuracy: 0.9750 - loss: 0.4460 - val_accuracy: 0.9947 - val_loss: 0.3689\n",
"Epoch 79/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9643 - loss: 0.4710\n",
"Epoch 79: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 71ms/step - accuracy: 0.9674 - loss: 0.4633 - val_accuracy: 0.9947 - val_loss: 0.3702\n",
"Epoch 80/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 48ms/step - accuracy: 0.9741 - loss: 0.4674\n",
"Epoch 80: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 67ms/step - accuracy: 0.9736 - loss: 0.4600 - val_accuracy: 0.9947 - val_loss: 0.3690\n",
"Epoch 81/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 53ms/step - accuracy: 0.9678 - loss: 0.4531\n",
"Epoch 81: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9714 - loss: 0.4485 - val_accuracy: 0.9947 - val_loss: 0.3692\n",
"Epoch 82/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 57ms/step - accuracy: 0.9779 - loss: 0.4602\n",
"Epoch 82: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 87ms/step - accuracy: 0.9776 - loss: 0.4540 - val_accuracy: 0.9947 - val_loss: 0.3694\n",
"Epoch 83/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 56ms/step - accuracy: 0.9658 - loss: 0.4708\n",
"Epoch 83: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 76ms/step - accuracy: 0.9711 - loss: 0.4630 - val_accuracy: 0.9947 - val_loss: 0.3680\n",
"Epoch 84/85\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 54ms/step - accuracy: 0.9782 - loss: 0.4492\n",
"Epoch 84: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9770 - loss: 0.4565 - val_accuracy: 0.9947 - val_loss: 0.3693\n",
"Epoch 85/85\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 56ms/step - accuracy: 0.9662 - loss: 0.4719\n",
"Epoch 85: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 78ms/step - accuracy: 0.9716 - loss: 0.4616 - val_accuracy: 0.9947 - val_loss: 0.3699\n"
]
}
],
"source": [
"img_size = IMG_SIZE # (224, 224)\n",
"\n",
"train_ds_full = tf.keras.utils.image_dataset_from_directory(\n",
" train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=img_size)\n",
"val_ds_full = tf.keras.utils.image_dataset_from_directory(\n",
" val_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=img_size)\n",
"\n",
"train_ds_full = (train_ds_full.map(augment_and_mix, num_parallel_calls=AUTOTUNE)\n",
" .map(preprocess_fn, num_parallel_calls=AUTOTUNE)\n",
" .map(add_sample_weight, num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"val_ds_full = (val_ds_full.map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)\n",
" .map(lambda img, lbl: (img, tf.one_hot(lbl, NUM_CLASSES)), num_parallel_calls=AUTOTUNE).prefetch(AUTOTUNE))\n",
"\n",
"# Full unfreeze\n",
"for layer in base_model.layers:\n",
" layer.trainable = True\n",
"\n",
"EPOCHS_P3 = 30\n",
"card_p3 = tf.data.experimental.cardinality(train_ds_full).numpy()\n",
"steps_p3 = card_p3 if card_p3 > 0 else 100\n",
"total_p3 = steps_p3 * EPOCHS_P3\n",
"warmup_p3 = steps_p3 * 2\n",
"\n",
"lr_schedule_p3 = WarmupCosineDecay(warmup_p3, total_p3, peak_lr=1e-4, min_lr=1e-7)\n",
"\n",
"model.compile(\n",
" optimizer=AdamW(\n",
" learning_rate=lr_schedule_p3, weight_decay=1e-4),\n",
" loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True, label_smoothing=0.10),\n",
" metrics=['accuracy']\n",
")\n",
"\n",
"print(\"Phase 3: Full fine-tuning at 224×
"history_3 = model.fit(train_ds_full, validation_data=val_ds_full,\n",
" epochs=EPOCHS_P1 + EPOCHS_P2 + EPOCHS_P3, initial_epoch=(history_2.epoch[-1] + 1) if history_2.epoch else EPOCHS_P1 + EPOCHS_P2,\n",
" callbacks=make_callbacks())"
]
2026-06-12 13:49:18 +00:00
},
{
"cell_type": "markdown",
"id": "6df4ef22",
"metadata": {},
"source": [
"## 12. SWA — Stochastic Weight Averaging\n",
"\n",
"15 epoch tambahan dengan cyclic LR (1e-5). SWA mengakumulasi rata-rata bobot\n",
"untuk menghasilkan **wider optima** — generalisasi lebih baik ke data out-of-distribution.\n",
"\n",
"Setelah SWA selesai, bobot SWA diterapkan kembali ke model.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 13,
2026-06-12 13:49:18 +00:00
"id": "290f2345",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"SWA: 15 epochs starting at epoch 86...\n",
"\n",
"SWA: starting weight averaging at epoch 86\n",
"Epoch 86/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 359ms/step - accuracy: 0.9738 - loss: 0.4644\n",
"Epoch 86: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m123s\u001b[0m 458ms/step - accuracy: 0.9748 - loss: 0.4571 - val_accuracy: 0.9947 - val_loss: 0.3677\n",
"Epoch 87/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9741 - loss: 0.4651\n",
"Epoch 87: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 74ms/step - accuracy: 0.9748 - loss: 0.4565 - val_accuracy: 0.9947 - val_loss: 0.3696\n",
"Epoch 88/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 55ms/step - accuracy: 0.9735 - loss: 0.4576\n",
"Epoch 88: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 72ms/step - accuracy: 0.9762 - loss: 0.4529 - val_accuracy: 0.9947 - val_loss: 0.3684\n",
"Epoch 89/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9681 - loss: 0.4679\n",
"Epoch 89: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 75ms/step - accuracy: 0.9719 - loss: 0.4654 - val_accuracy: 0.9947 - val_loss: 0.3695\n",
"Epoch 90/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9753 - loss: 0.4659\n",
"Epoch 90: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 72ms/step - accuracy: 0.9767 - loss: 0.4555 - val_accuracy: 0.9947 - val_loss: 0.3697\n",
"Epoch 91/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 58ms/step - accuracy: 0.9782 - loss: 0.4525\n",
"Epoch 91: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 76ms/step - accuracy: 0.9773 - loss: 0.4555 - val_accuracy: 0.9947 - val_loss: 0.3684\n",
"Epoch 92/100\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 58ms/step - accuracy: 0.9730 - loss: 0.4624\n",
"Epoch 92: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 76ms/step - accuracy: 0.9773 - loss: 0.4534 - val_accuracy: 0.9947 - val_loss: 0.3691\n",
"Epoch 93/100\n",
"\u001b[1m109/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9726 - loss: 0.4550\n",
"Epoch 93: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 68ms/step - accuracy: 0.9739 - loss: 0.4524 - val_accuracy: 0.9947 - val_loss: 0.3701\n",
"Epoch 94/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 51ms/step - accuracy: 0.9751 - loss: 0.4596\n",
"Epoch 94: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9759 - loss: 0.4537 - val_accuracy: 0.9947 - val_loss: 0.3686\n",
"Epoch 95/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 55ms/step - accuracy: 0.9763 - loss: 0.4528\n",
"Epoch 95: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 74ms/step - accuracy: 0.9793 - loss: 0.4439 - val_accuracy: 0.9947 - val_loss: 0.3688\n",
"Epoch 96/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9757 - loss: 0.4581\n",
"Epoch 96: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 70ms/step - accuracy: 0.9773 - loss: 0.4437 - val_accuracy: 0.9947 - val_loss: 0.3672\n",
"Epoch 97/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 52ms/step - accuracy: 0.9660 - loss: 0.4680\n",
"Epoch 97: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 73ms/step - accuracy: 0.9697 - loss: 0.4596 - val_accuracy: 0.9947 - val_loss: 0.3704\n",
"Epoch 98/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 57ms/step - accuracy: 0.9780 - loss: 0.4543\n",
"Epoch 98: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 76ms/step - accuracy: 0.9765 - loss: 0.4490 - val_accuracy: 0.9947 - val_loss: 0.3706\n",
"Epoch 99/100\n",
"\u001b[1m110/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 50ms/step - accuracy: 0.9711 - loss: 0.4666\n",
"Epoch 99: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 68ms/step - accuracy: 0.9745 - loss: 0.4585 - val_accuracy: 0.9947 - val_loss: 0.3691\n",
"Epoch 100/100\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 57ms/step - accuracy: 0.9722 - loss: 0.4747\n",
"Epoch 100: val_accuracy did not improve from 0.99602\n",
"\u001b[1m111/111\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m10s\u001b[0m 77ms/step - accuracy: 0.9748 - loss: 0.4610 - val_accuracy: 0.9947 - val_loss: 0.3687\n",
"SWA weights applied (15 models averaged).\n",
"SWA complete. Final model has SWA weights applied.\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"EPOCHS_SWA = 15\n",
"swa_start_epoch = (history_3.epoch[-1] + 1) if history_3.epoch else EPOCHS_P1 + EPOCHS_P2 + EPOCHS_P3\n",
"\n",
"swa_cb = SWACallback(start_epoch=swa_start_epoch, swa_lr=1e-5)\n",
"\n",
"model.compile(\n",
" optimizer=AdamW(\n",
" learning_rate=1e-5, weight_decay=1e-4),\n",
" loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True, label_smoothing=0.10),\n",
" metrics=['accuracy']\n",
")\n",
"\n",
"print(f\"SWA: {EPOCHS_SWA} epochs starting at epoch {swa_start_epoch + 1}...\")\n",
"history_swa = model.fit(train_ds_full, validation_data=val_ds_full,\n",
" epochs=swa_start_epoch + EPOCHS_SWA, initial_epoch=swa_start_epoch,\n",
" callbacks=make_callbacks() + [swa_cb])\n",
"\n",
"# Apply SWA weights\n",
"swa_cb.apply_swa_weights()\n",
"print(f\"SWA complete. Final model has SWA weights applied.\")\n"
]
},
{
"cell_type": "markdown",
"id": "2d889f2a",
"metadata": {},
"source": [
"## 13. Plot Training History (Gabungan Semua Fase)\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 16,
2026-06-12 13:49:18 +00:00
"id": "41e79742",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"data": {
"image/png": "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
"text/plain": [
"<Figure size 1600x600 with 2 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
2026-06-12 13:49:18 +00:00
"source": [
"# Gabungkan semua history\n",
"acc = (history_1.history['accuracy'] + history_2.history['accuracy'] +\n",
" history_3.history['accuracy'] + history_swa.history['accuracy'])\n",
"val_acc = (history_1.history['val_accuracy'] + history_2.history['val_accuracy'] +\n",
" history_3.history['val_accuracy'] + history_swa.history['val_accuracy'])\n",
"loss = (history_1.history['loss'] + history_2.history['loss'] +\n",
" history_3.history['loss'] + history_swa.history['loss'])\n",
"val_loss = (history_1.history['val_loss'] + history_2.history['val_loss'] +\n",
" history_3.history['val_loss'] + history_swa.history['val_loss'])\n",
"\n",
"b1 = len(history_1.history['accuracy']) - 1\n",
"b2 = b1 + len(history_2.history['accuracy'])\n",
"b3 = b2 + len(history_3.history['accuracy'])\n",
"\n",
"plt.figure(figsize=(16, 6))\n",
"plt.subplot(1, 2, 1)\n",
"plt.plot(acc, label='Training Accuracy', linewidth=2)\n",
"plt.plot(val_acc, label='Validation Accuracy', linewidth=2)\n",
"plt.axvline(x=b1, color='gray', linestyle='--', alpha=0.7, label='P2 (192)')\n",
"plt.axvline(x=b2, color='black', linestyle='--', alpha=0.7, label='P3 (224)')\n",
"plt.axvline(x=b3, color='blue', linestyle='--', alpha=0.7, label='SWA start')\n",
"plt.legend(fontsize=10)\n",
"plt.title('Training & Validation Accuracy', fontsize=14)\n",
"plt.xlabel('Epoch'); plt.ylabel('Accuracy'); plt.grid(alpha=0.3)\n",
"\n",
"plt.subplot(1, 2, 2)\n",
"plt.plot(loss, label='Training Loss', linewidth=2)\n",
"plt.plot(val_loss, label='Validation Loss', linewidth=2)\n",
"plt.axvline(x=b1, color='gray', linestyle='--', alpha=0.7, label='P2 (192)')\n",
"plt.axvline(x=b2, color='black', linestyle='--', alpha=0.7, label='P3 (224)')\n",
"plt.axvline(x=b3, color='blue', linestyle='--', alpha=0.7, label='SWA start')\n",
"plt.legend(fontsize=10)\n",
"plt.title('Training & Validation Loss', fontsize=14)\n",
"plt.xlabel('Epoch'); plt.ylabel('Loss'); plt.grid(alpha=0.3)\n",
"\n",
"plt.tight_layout()\n",
"plt.show()\n"
]
},
{
"cell_type": "markdown",
"id": "27b920dd",
"metadata": {},
"source": [
"## 14. Temperature Scaling — Confidence Calibration\n",
"\n",
"Model deep learning cenderung **overconfident** — softmax probability tinggi tapi tidak mencerminkan\n",
"akurasi sebenarnya. Temperature scaling mengoptimalkan parameter T pada validation set:\n",
"\n",
"$$P_{calibrated} = softmax(logits / T)$$\n",
"\n",
"T > 1 → distribusi lebih flat (less confident).\n",
"T < 1 → distribusi lebih tajam (more confident).\n",
"T = 1 → tidak berubah (default).\n",
"\n",
"ECE (Expected Calibration Error) mengukur seberapa baik confidence sesuai dengan akurasi.\n",
"Target: **ECE < 0.05** setelah temperature scaling.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 17,
2026-06-12 13:49:18 +00:00
"id": "d79dbb47",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Collecting validation logits...\n",
"ECE before scaling (T=1.0): 0.0661\n",
"Optimal temperature: T = 0.5303\n",
"ECE after scaling (T=0.5303): 0.0031\n",
"Calibration metadata saved to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/calibration.json\n"
]
},
{
"data": {
"image/png": "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
"text/plain": [
"<Figure size 1200x500 with 2 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
2026-06-12 13:49:18 +00:00
"source": [
"def compute_ece(probs, true_labels, n_bins=15):\n",
" # Expected Calibration Error.\n",
" confs = np.max(probs, axis=1)\n",
" preds = np.argmax(probs, axis=1)\n",
" true = np.argmax(true_labels, axis=1)\n",
" accs = (preds == true).astype(np.float32)\n",
" bins = np.linspace(0, 1, n_bins + 1)\n",
" ece = 0.0\n",
" bin_stats = []\n",
" for i in range(n_bins):\n",
" in_bin = (confs > bins[i]) & (confs <= bins[i + 1])\n",
" n = np.sum(in_bin)\n",
" if n > 0:\n",
" bin_acc = np.mean(accs[in_bin])\n",
" bin_conf = np.mean(confs[in_bin])\n",
" ece += (n / len(confs)) * np.abs(bin_acc - bin_conf)\n",
" bin_stats.append((bins[i], n, bin_acc, bin_conf))\n",
" return ece, bin_stats\n",
"\n",
"# Collect logits and labels from validation set\n",
"print(\"Collecting validation logits...\")\n",
"logits_model = tf.keras.Model(model.input, model.output)\n",
"\n",
"all_logits = []\n",
"all_labels = []\n",
"for images, labels in val_ds_full.unbatch().batch(BATCH_SIZE):\n",
" all_logits.append(logits_model.predict_on_batch(images))\n",
" all_labels.append(labels.numpy())\n",
"\n",
"all_logits = np.concatenate(all_logits, axis=0)\n",
"all_labels = np.concatenate(all_labels, axis=0)\n",
"\n",
"# ECE before scaling (T=1)\n",
"probs_raw = tf.nn.softmax(all_logits).numpy()\n",
"ece_raw, _ = compute_ece(probs_raw, all_labels)\n",
"print(f\"ECE before scaling (T=1.0): {ece_raw:.4f}\")\n",
"\n",
"# Optimize T on validation set\n",
"if HAS_SCIPY:\n",
" def nll_temperature(T):\n",
" scaled = all_logits / float(T)\n",
" probs = tf.nn.softmax(scaled).numpy()\n",
" probs = np.clip(probs, 1e-7, 1.0 - 1e-7)\n",
" return -np.mean(np.log(np.sum(all_labels * probs, axis=1)))\n",
"\n",
" result = minimize_scalar(nll_temperature, bounds=(0.1, 5.0), method='bounded')\n",
" T_opt = result.x\n",
" print(f\"Optimal temperature: T = {T_opt:.4f}\")\n",
"else:\n",
" # Grid search fallback\n",
" best_nll, T_opt = float('inf'), 1.0\n",
" for T in np.linspace(0.5, 4.0, 36):\n",
" scaled = all_logits / T\n",
" probs = tf.nn.softmax(scaled).numpy()\n",
" probs = np.clip(probs, 1e-7, 1.0 - 1e-7)\n",
" nll = -np.mean(np.log(np.sum(all_labels * probs, axis=1)))\n",
" if nll < best_nll:\n",
" best_nll = nll\n",
" T_opt = T\n",
" print(f\"Optimal temperature (grid): T = {T_opt:.4f}\")\n",
"\n",
"# ECE after scaling\n",
"probs_cal = tf.nn.softmax(all_logits / T_opt).numpy()\n",
"ece_cal, bin_stats = compute_ece(probs_cal, all_labels)\n",
"print(f\"ECE after scaling (T={T_opt:.4f}): {ece_cal:.4f}\")\n",
"\n",
"# Save calibration metadata\n",
"calibration_meta = {\n",
" \"temperature\": float(T_opt),\n",
" \"conf_threshold_high\": 0.70,\n",
" \"conf_threshold_low\": 0.45,\n",
" \"ece_raw\": float(ece_raw),\n",
" \"ece_calibrated\": float(ece_cal),\n",
"}\n",
"with open(os.path.join(ckpt_dir, \"calibration.json\"), \"w\") as f:\n",
" json.dump(calibration_meta, f, indent=2)\n",
"print(f\"Calibration metadata saved to {os.path.join(ckpt_dir, 'calibration.json')}\")\n",
"\n",
"# Reliability diagram\n",
2026-06-12 15:00:15 +00:00
"n_bins = 15\n",
2026-06-12 13:49:18 +00:00
"plt.figure(figsize=(12, 5))\n",
"\n",
"plt.subplot(1, 2, 1)\n",
"if bin_stats:\n",
" bin_mids = [(s[0] + s[0] + 1/n_bins)/2 for s in bin_stats]\n",
" bin_accs = [s[2] for s in bin_stats]\n",
" bin_confs = [s[3] for s in bin_stats]\n",
" plt.bar(bin_mids, bin_accs, width=0.05, alpha=0.5, label='Accuracy')\n",
" plt.bar(bin_mids, bin_confs, width=0.05, alpha=0.3, label='Confidence')\n",
"plt.plot([0, 1], [0, 1], 'k--', alpha=0.3)\n",
"plt.xlabel('Confidence'); plt.ylabel('Accuracy')\n",
"plt.title(f'Reliability Diagram (T={T_opt:.2f})')\n",
"plt.legend(); plt.grid(alpha=0.3)\n",
"\n",
"plt.subplot(1, 2, 2)\n",
"conf_raw = np.max(probs_raw, axis=1)\n",
"conf_cal = np.max(probs_cal, axis=1)\n",
"plt.hist(conf_raw, bins=30, alpha=0.5, label='Before scaling', density=True)\n",
"plt.hist(conf_cal, bins=30, alpha=0.5, label='After scaling', density=True)\n",
"plt.xlabel('Max Confidence'); plt.ylabel('Density')\n",
"plt.title('Confidence Distribution')\n",
"plt.legend(); plt.grid(alpha=0.3)\n",
"\n",
"plt.tight_layout()\n",
2026-06-12 15:00:15 +00:00
"plt.show()"
2026-06-12 13:49:18 +00:00
]
},
{
"cell_type": "markdown",
"id": "6e4a717b",
"metadata": {},
"source": [
"## 15. Test Time Augmentation + Evaluasi\n",
"\n",
"Menggunakan TTA 5×
"Model output adalah **raw logits** → temperature scaling → softmax.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 20,
2026-06-12 13:49:18 +00:00
"id": "beb96b8b",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Found 758 files belonging to 4 classes.\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2026-06-12 14:58:18.264137: I tensorflow/core/framework/local_rendezvous.cc:407] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence\n",
"2026-06-12 14:58:29.241715: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2424', 252 bytes spill stores, 252 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.245951: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2424', 16 bytes spill stores, 16 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.395233: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2531', 252 bytes spill stores, 252 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.456816: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2531_0', 492 bytes spill stores, 1464 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.498326: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2531', 16 bytes spill stores, 16 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.825599: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2424', 5912 bytes spill stores, 5980 bytes spill loads\n",
"\n",
"2026-06-12 14:58:29.891163: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2424', 5356 bytes spill stores, 5344 bytes spill loads\n",
"\n",
"2026-06-12 14:58:30.050037: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2531', 5356 bytes spill stores, 5344 bytes spill loads\n",
"\n",
"2026-06-12 14:58:30.128273: I external/local_xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_2531', 5912 bytes spill stores, 5980 bytes spill loads\n",
"\n",
"2026-06-12 14:58:31.572783: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:31.729121: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:32.521772: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:32.685014: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:33.300056: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:33.451394: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:33.822116: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:33.989295: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:34.912339: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:35.068434: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:35.265910: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n",
"2026-06-12 14:58:35.424360: E external/local_xla/xla/stream_executor/cuda/cuda_timer.cc:86] Delay kernel timed out: measured time has sub-optimal accuracy. There may be a missing warmup execution, please investigate in Nsight Systems.\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
" TTA step 1/5\n",
" TTA step 2/5\n",
" TTA step 3/5\n",
" TTA step 4/5\n",
" TTA step 5/5\n",
"\n",
"Test Accuracy (TTA 5x, T=0.53): 99.34%\n",
"ECE on test set: 0.0028\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
2026-06-12 15:00:15 +00:00
"# Gunakan model dari memory (sudah SWA weights applied dari Cell 12)\n",
"# Tidak perlu load checkpoint — model sudah siap di kernel.\n",
"best_model = model\n",
2026-06-12 13:49:18 +00:00
"\n",
"# Collect test images\n",
"raw_test_ds = tf.keras.utils.image_dataset_from_directory(\n",
" test_dir, shuffle=False, batch_size=BATCH_SIZE, image_size=IMG_SIZE)\n",
"\n",
"test_images = []\n",
"test_labels_raw = []\n",
"for images, labels in raw_test_ds.unbatch():\n",
" test_images.append(images.numpy())\n",
" test_labels_raw.append(labels.numpy())\n",
"\n",
"test_images = np.array(test_images)\n",
"test_labels_true = tf.one_hot(np.array(test_labels_raw), NUM_CLASSES).numpy()\n",
"\n",
"# ── TTA 5x ──\n",
"TTA_STEPS = 5\n",
"tta_logits = []\n",
"\n",
"for i in range(TTA_STEPS):\n",
" aug_images = geo_aug(test_images, training=True)\n",
" aug_images = preprocess_input(aug_images)\n",
" logits = best_model.predict(aug_images, batch_size=BATCH_SIZE, verbose=0)\n",
" tta_logits.append(logits)\n",
" print(f\" TTA step {i+1}/{TTA_STEPS}\")\n",
"\n",
"mean_logits = np.mean(tta_logits, axis=0)\n",
"# Apply temperature scaling\n",
"mean_cal_probs = tf.nn.softmax(mean_logits / T_opt).numpy()\n",
"\n",
"test_preds = np.argmax(mean_cal_probs, axis=1)\n",
"test_true = np.argmax(test_labels_true, axis=1)\n",
"tta_acc = np.mean(test_preds == test_true)\n",
"\n",
"print(f\"\\nTest Accuracy (TTA {TTA_STEPS}x, T={T_opt:.2f}): {tta_acc*100:.2f}%\")\n",
"\n",
"# ECE on test set\n",
"ece_test, _ = compute_ece(mean_cal_probs, test_labels_true)\n",
2026-06-12 15:00:15 +00:00
"print(f\"ECE on test set: {ece_test:.4f}\")"
2026-06-12 13:49:18 +00:00
]
},
{
"cell_type": "markdown",
"id": "b6bb88ab",
"metadata": {},
"source": [
"## 16. Per-Source Accuracy Breakdown\n",
"\n",
"Mengukur akurasi per source prefix untuk mendeteksi **domain gap**.\n",
"Source yang akurasinya collapse (< 70%) menunjukkan model belum robust.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 21,
2026-06-12 13:49:18 +00:00
"id": "a24e9a14",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Sources found: ['Lab_Corn', 'Phone']\n",
"Source Count Accuracy\n",
"--------------------------------------\n",
"Lab_Corn 458 99.1%\n",
"Phone 300 99.7%\n",
"--------------------------------------\n",
"ALL 758 99.3%\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
"def get_source(filename):\n",
" f = os.path.splitext(filename)[0]\n",
" if f.startswith('IMG_'): return 'Phone'\n",
" if f.startswith('Corn_'): return 'Lab_Corn'\n",
" for p in ['CBS', 'GLS', 'NLS', 'CLS']:\n",
" if f.startswith(p): return 'Lab_Disease'\n",
" for p in ['SCR', 'CR', 'NLB', 'SLB', 'SRS']:\n",
" if f.startswith(p): return 'Lab_RustBlight'\n",
" return 'Other'\n",
"\n",
"# Map each test image to its source\n",
"test_files = []\n",
"for cn in class_names:\n",
" cp = os.path.join(test_dir, cn)\n",
" if os.path.isdir(cp):\n",
" test_files.extend([(f, cn, get_source(f)) for f in sorted(os.listdir(cp))])\n",
"\n",
"sources = set(s for _, _, s in test_files)\n",
"print(f\"Sources found: {sorted(sources)}\")\n",
"print(f\"{'Source':<18} {'Count':>6} {'Accuracy':>10}\")\n",
"print(\"-\" * 38)\n",
"\n",
"for src in sorted(sources):\n",
" indices = [i for i, (_, _, s) in enumerate(test_files) if s == src]\n",
" if not indices: continue\n",
" n = len(indices)\n",
" acc = np.mean(test_preds[indices] == test_true[indices])\n",
" print(f\"{src:<18} {n:>6} {acc*100:>9.1f}%\")\n",
"\n",
"# Overall with count\n",
"overall_acc = np.mean(test_preds == test_true)\n",
"print(\"-\" * 38)\n",
"print(f\"{'ALL':<18} {len(test_preds):>6} {overall_acc*100:>9.1f}%\")\n"
]
},
{
"cell_type": "markdown",
"id": "52754df5",
"metadata": {},
"source": [
"## 17. Classification Report & Confusion Matrix\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 22,
2026-06-12 13:49:18 +00:00
"id": "1a5b2c96",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"Classification Report:\n",
"\n",
" precision recall f1-score support\n",
"\n",
" Bercak Daun 0.98 0.98 0.98 87\n",
" Daun Sehat 1.00 1.00 1.00 325\n",
" Hawar Daun 0.99 1.00 1.00 150\n",
" Karat Daun 0.99 0.99 0.99 196\n",
"\n",
" accuracy 0.99 758\n",
" macro avg 0.99 0.99 0.99 758\n",
"weighted avg 0.99 0.99 0.99 758\n",
"\n"
]
},
{
"data": {
"image/png": "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
"text/plain": [
"<Figure size 1000x800 with 2 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
2026-06-12 13:49:18 +00:00
"source": [
"print(\"\\nClassification Report:\\n\")\n",
"print(classification_report(test_true, test_preds, target_names=class_names))\n",
"\n",
"cm = confusion_matrix(test_true, test_preds)\n",
"plt.figure(figsize=(10, 8))\n",
"sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',\n",
" xticklabels=class_names, yticklabels=class_names)\n",
"plt.title(f'Confusion Matrix (TTA {TTA_STEPS}x)', fontsize=14)\n",
"plt.xlabel('Predicted', fontsize=12)\n",
"plt.ylabel('True', fontsize=12)\n",
"plt.tight_layout()\n",
"plt.show()\n"
]
},
{
"cell_type": "markdown",
"id": "528ef53b",
"metadata": {},
"source": [
"## 18. Simpan Model & Calibration Metadata\n",
"\n",
"Menyimpan model final (dengan SWA weights) dan calibration metadata ke `model/`.\n"
]
},
{
"cell_type": "code",
2026-06-12 15:00:15 +00:00
"execution_count": 23,
2026-06-12 13:49:18 +00:00
"id": "292fef1a",
"metadata": {},
2026-06-12 15:00:15 +00:00
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Model saved to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras\n",
"Weights saved to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/model.weights.h5\n",
"Labels + calibration metadata saved to model/labels.json\n",
"Classes: ['Bercak Daun', 'Daun Sehat', 'Hawar Daun', 'Karat Daun']\n",
"Temperature: 0.5303\n"
]
}
],
2026-06-12 13:49:18 +00:00
"source": [
2026-06-12 15:00:15 +00:00
"# Save final model (keras format — mungkin tidak bisa di-load karena Lambda)\n",
2026-06-12 13:49:18 +00:00
"model.save(final_path := os.path.join(ckpt_dir, 'best_model.keras'))\n",
2026-06-12 15:00:15 +00:00
"# Save weights H5 (portable — used by save_model.py for export)\n",
"model.save_weights(os.path.join(ckpt_dir, 'model.weights.h5'))\n",
2026-06-12 13:49:18 +00:00
"print(f\"Model saved to {final_path}\")\n",
2026-06-12 15:00:15 +00:00
"print(f\"Weights saved to {os.path.join(ckpt_dir, 'model.weights.h5')}\")\n",
2026-06-12 13:49:18 +00:00
"\n",
"# Copy calibration metadata to model directory\n",
"model_export_dir = os.path.join(os.getcwd(), 'model')\n",
"os.makedirs(model_export_dir, exist_ok=True)\n",
"\n",
"# Export labels.json with calibration metadata\n",
"cal_path = os.path.join(ckpt_dir, 'calibration.json')\n",
"if os.path.exists(cal_path):\n",
" with open(cal_path) as f:\n",
" cal_meta = json.load(f)\n",
"\n",
" labels_json = {\n",
" \"version\": \"3.0\",\n",
" \"labels\": class_names,\n",
" \"temperature\": cal_meta[\"temperature\"],\n",
" \"conf_threshold_high\": cal_meta[\"conf_threshold_high\"],\n",
" \"conf_threshold_low\": cal_meta[\"conf_threshold_low\"],\n",
" \"input_size\": [224, 224],\n",
" \"input_range\": [0, 255],\n",
" \"preprocessing\": \"resize_bilinear_224x224_no_normalization\",\n",
" \"architecture\": \"EfficientNetV2B0 + CBAM + Dense(512)\",\n",
" \"output_type\": \"logits\",\n",
" }\n",
"else:\n",
" labels_json = {\n",
" \"version\": \"3.0\",\n",
" \"labels\": class_names,\n",
" \"temperature\": 1.0,\n",
" \"input_size\": [224, 224],\n",
" \"input_range\": [0, 255],\n",
" \"output_type\": \"logits\",\n",
" }\n",
"\n",
"with open(os.path.join(model_export_dir, 'labels.json'), 'w') as f:\n",
" json.dump(labels_json, f, indent=2)\n",
"print(\"Labels + calibration metadata saved to model/labels.json\")\n",
"\n",
"# Export class names list (legacy)\n",
"with open(os.path.join(model_export_dir, 'labels.json'), 'r') as f:\n",
" pass # already written above\n",
"print(f\"Classes: {class_names}\")\n",
2026-06-12 15:00:15 +00:00
"print(f\"Temperature: {labels_json['temperature']:.4f}\")"
2026-06-12 13:49:18 +00:00
]
},
{
"cell_type": "markdown",
"id": "62675ad6",
"metadata": {},
"source": [
"## 19. Export Model untuk Produksi\n",
"\n",
"Setelah training selesai, jalankan pipeline ekspor secara berurutan:\n",
"\n",
"### 1. SavedModel + TFLite\n",
"```bash\n",
"python save_model.py\n",
"```\n",
"Memuat `best_model/best_model.keras`, membangun arsitektur bersih, dan mengekspor ke:\n",
"- `model/saved_model/` — format produksi (output **raw logits**)\n",
"- `model/model.tflite` — untuk perangkat mobile/edge (INT8 quantization)\n",
"\n",
"### 2. ONNX (Rust ML Service)\n",
"```bash\n",
"python convert_onnx.py\n",
"```\n",
"Mengonversi SavedModel ke `model/model.onnx` untuk Rust/Axum/ONNX Runtime.\n",
"Output: **raw logits** (softmax + temperature scaling di Rust service).\n",
"\n",
"### 3. TensorFlow.js (Web) — optional\n",
"```bash\n",
"export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python\n",
"tensorflowjs_converter --input_format=tf_saved_model --output_format=tfjs_graph_model --signature_name=serving_default --saved_model_tags=serve model/saved_model model/tfjs_model\n",
"```\n",
"\n",
"> **Catatan v3.0**: Model output adalah **raw logits** (tanpa softmax). Rust service menerapkan\n",
"> temperature scaling: `softmax(logits / T)` dengan `T` dari `labels.json`.\n",
"> Status prediksi ditentukan dari confidence: confident (≥70%), uncertain (45-70%), rejected (<45%).\n"
]
},
{
"cell_type": "markdown",
"id": "de8c53b7",
"metadata": {},
"source": [
"## 20. Model Card — ZeaVis Edu v3.0\n",
"\n",
"| Atribut | Detail |\n",
"|---|---|\n",
"| **Nama Model** | ZeaVis Edu v3.0 — CBAM + RandAugment Classifier |\n",
"| **Versi** | 3.0 |\n",
"| **Arsitektur** | EfficientNetV2B0 + CBAM Attention + GAP + Dense(512) |\n",
"| **Params** | ~6.6M (5.9M base + 0.7M head) — 2×
"| **Framework** | TensorFlow 2.x / Keras (float32) |\n",
"| **Output** | Raw logits → temperature scaling → softmax |\n",
"| **Dataset** | ~6000 gambar (4 kelas) — stratified split by source |\n",
"| **Kelas** | Bercak Daun, Daun Sehat, Hawar Daun, Karat Daun |\n",
"| **Input** | RGB 224×
"| **Augmentasi** | RandAugment (15 ops, N=3) + Fog + Shadow + MixUp + CutMix + RandomErasing |\n",
"| **Training** | Progressive 128→160→192→224 + CosineDecay + SWA |\n",
"| **Calibration** | Temperature scaling (T optimized on val), ECE target < 0.05 |\n",
"| **Decision** | Confident (≥0.70), Uncertain (0.45-0.70), Rejected (<0.45) |\n",
"| **Target** | ≥90% real-world accuracy, per-source accuracy ≥70% semua domain |\n",
"| **Etika** | Hanya untuk edukasi/penelitian pertanian. Bukan pengganti diagnosis ahli. |\n",
"\n"
]
},
{
"cell_type": "markdown",
"id": "f431a07c",
"metadata": {},
"source": [
"## 21. Upload ke Hugging Face\n",
"\n",
"```python\n",
"from huggingface_hub import HfApi\n",
"api = HfApi()\n",
"api.upload_folder(\n",
" folder_path=\"model\",\n",
" repo_id=\"zeavis-edu/corn-leaf-disease-classifier\",\n",
" repo_type=\"model\",\n",
")\n",
"```\n",
"Upload model SavedModel, TFLite, ONNX, TF.js, dan metadata ke HF Hub.\n"
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.3"
}
},
"nbformat": 4,
"nbformat_minor": 5
2026-06-12 15:00:15 +00:00
}
