MythEclipse/zeavis-edu

Fork 0

Files

T

MythEclipse 8137057186 a

2026-06-11 19:42:02 +00:00

1.9 MiB

Raw Blame History

ZeaVis Edu — Corn Leaf Disease Classifier

Mengklasifikasikan penyakit daun jagung (Bercak Daun, Hawar Daun, Karat Daun, Daun Sehat) menggunakan EfficientNetV2B0 dengan transfer learning.

1. Persiapan Lingkungan

Mengimpor pustaka, mengatur seed, dan mengoptimalkan konfigurasi. Presisi float32 digunakan untuk komputasi. Resolusi gambar: 224x224 (EfficientNetV2B0).

In [12]:

!pip install -r requirements.txt

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)
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)
Requirement already satisfied: gdown in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 3)) (6.1.0)
Requirement already satisfied: numpy in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 4)) (2.1.3)
Requirement already satisfied: matplotlib in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 5)) (3.10.9)
Requirement already satisfied: seaborn in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 6)) (0.13.2)
Requirement already satisfied: pillow in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 7)) (12.2.0)
Requirement already satisfied: split-folders in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 8)) (0.6.1)
Requirement already satisfied: scikit-learn in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 9)) (1.9.0)
Requirement already satisfied: tf2onnx in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 10)) (1.17.0)
Requirement already satisfied: onnxruntime in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from -r requirements.txt (line 11)) (1.26.0)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
Requirement already satisfied: contourpy>=1.0.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (1.3.3)
Requirement already satisfied: cycler>=0.10 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (0.12.1)
Requirement already satisfied: fonttools>=4.22.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (4.63.0)
Requirement already satisfied: kiwisolver>=1.3.1 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (1.5.0)
Requirement already satisfied: pyparsing>=3 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (3.3.2)
Requirement already satisfied: python-dateutil>=2.7 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from matplotlib->-r requirements.txt (line 5)) (2.9.0.post0)
Requirement already satisfied: pandas>=1.2 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from seaborn->-r requirements.txt (line 6)) (3.0.3)
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 9)) (1.17.1)
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 9)) (1.5.3)
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 9)) (2.22.1)
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 9)) (3.6.0)
Requirement already satisfied: onnx>=1.14.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from tf2onnx->-r requirements.txt (line 10)) (1.21.0)
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)
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)
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)
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)
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)
Requirement already satisfied: rich>=11.1 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)) (15.0.0)
Requirement already satisfied: PyYAML>=5.4.1 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)) (6.0.3)
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)
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)
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)
Requirement already satisfied: markdown-it-py>=2.2.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from rich>=11.1->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (4.2.0)
Requirement already satisfied: pygments<3.0.0,>=2.13.0 in /home/asephs/ZeaVis-Edu/.venv/lib/python3.12/site-packages (from rich>=11.1->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (2.20.0)
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>=11.1->flax>=0.7.2->tensorflowjs==4.22.0->-r requirements.txt (line 2)) (0.1.2)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
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)
Requirement already satisfied: fsspec 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)) (2026.4.0)
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)
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)

In [13]:

import os
import shutil
import zipfile
import random
from collections import Counter
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import splitfolders
from PIL import Image

import tensorflow as tf
from tensorflow.keras import layers, models, callbacks
from tensorflow.keras.applications import EfficientNetV2B0
from tensorflow.keras.applications.efficientnet_v2 import preprocess_input
from tensorflow.keras.optimizers import AdamW
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.utils.class_weight import compute_class_weight

# Detect environment
try:
    from google.colab import drive
    IS_COLAB = True
    print("Running on Google Colab")
except ModuleNotFoundError:
    IS_COLAB = False
    print(f"Running locally (TF {tf.__version__}, GPU: {tf.config.list_physical_devices('GPU')})")

tf.keras.mixed_precision.set_global_policy('float32')

# Hyperparams
IMG_SIZE = (224, 224)
BATCH_SIZE = 32
SEED = 42
random.seed(SEED)
np.random.seed(SEED)
tf.random.set_seed(SEED)

AUTOTUNE = tf.data.AUTOTUNE
print(f"Setup OK. IMG={IMG_SIZE}, BATCH={BATCH_SIZE}")

Running locally (TF 2.19.0, GPU: [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')])
Setup OK. IMG=(224, 224), BATCH=32

2. Download dan Ekstraksi Dataset

In [14]:

if IS_COLAB:
    drive.mount('/content/drive')
    archive_path = '/content/drive/MyDrive/jagung/dataset_jagung.zip'
    destination_path = '/content/dataset_jagung.zip'
    extract_path = '/content/dataset'
else:
    import gdown
    base = os.getcwd()
    archive_path = os.path.join(base, 'dataset_jagung.zip')
    destination_path = archive_path
    extract_path = os.path.join(base, 'dataset')
    DRIVE_FILE_ID = "1s0H2lDOQVCixywk5eZXJz2i9jj4JihxJ"
    if not os.path.exists(archive_path):
        print("Downloading from Google Drive...")
        try:
            gdown.download(f"https://drive.google.com/uc?id={DRIVE_FILE_ID}", archive_path, quiet=False)
        except Exception as e:
            print(f"Download failed: {e}")

if os.path.exists(destination_path):
    if not os.path.exists(extract_path) or len(os.listdir(extract_path)) == 0:
        os.makedirs(extract_path, exist_ok=True)
        print("Extracting dataset...")
        try:
            with zipfile.ZipFile(destination_path, 'r') as zip_ref:
                zip_ref.extractall(path=extract_path)
            print("Extraction completed!")
        except Exception as e:
            print(f"Extraction failed: {e}")
    else:
        print("Dataset ready.")

Dataset ready.

3. Data Cleaning dan Validasi Gambar (RGB)

In [15]:

# --- Determine dataset path ---
# Colab: zip extracts to dataset/ with dataset_jagung_v1/ subfolder
# Local: zip extracts classes directly to dataset/
if IS_COLAB:
    dataset_path = "/content/dataset/dataset_jagung_v1"
else:
    candidate = os.path.join(extract_path, "dataset_jagung_v1")
    if os.path.isdir(candidate):
        dataset_path = candidate
    else:
        dataset_path = extract_path  # classes are directly in dataset/

print(f"Dataset path: {dataset_path}")

def clean_image_data(directory):
    removed_count = 0
    for root, dirs, files in os.walk(directory):
        for file in files:
            file_path = os.path.join(root, file)
            try:
                img = Image.open(file_path)
                img.verify()
                img = Image.open(file_path)
                if img.mode != 'RGB':
                    img = img.convert('RGB')
                    img.save(file_path)
            except Exception:
                print(f"Removing: {file_path}")
                os.remove(file_path)
                removed_count += 1
    return removed_count

print("Cleaning data...")
removed = clean_image_data(dataset_path)
print(f"Done. {removed} problematic files removed.")

Dataset path: /home/asephs/ZeaVis-Edu/Machine_Learning/dataset
Cleaning data...
Done. 0 problematic files removed.

4. Data Splitting (Train:Validation:Test = 70:15:15)

In [16]:

if IS_COLAB:
    output_dir = "/content/dataset_split"
else:
    output_dir = os.path.join(os.getcwd(), "dataset_split")

if os.path.exists(output_dir):
    shutil.rmtree(output_dir)

print("Splitting dataset 70:15:15...")
splitfolders.ratio(dataset_path, output=output_dir,
                   seed=SEED, ratio=(0.7, 0.15, 0.15),
                   group_prefix=None, move=False)
print("Done.")

train_dir = os.path.join(output_dir, 'train')
val_dir = os.path.join(output_dir, 'val')
test_dir = os.path.join(output_dir, 'test')

# Count
def count_images(path):
    return sum(len(files) for _, _, files in os.walk(path))

print(f'Train: {count_images(train_dir)} | Val: {count_images(val_dir)} | Test: {count_images(test_dir)}')

Splitting dataset 70:15:15...

Copying files: 6988 files [00:08, 830.89 files/s]

Done.
Train: 4890 | Val: 1046 | Test: 1052

5. Data Loader dengan Augmentasi Lanjutan

Menggunakan pipeline augmentasi berlapis:

Geometric: RandomFlip, Rotation, Zoom, Translation, Contrast, Brightness
CutMix & MixUp: mencampur dua gambar dan label (probabilitas 50% masing-masing)
RandomErasing: menghapus area acak pada gambar
GaussianNoise di dalam model (layer terpisah)
Class weights: menangani ketidakseimbangan kelas (capped max 2.5)

In [17]:

# --- Load datasets ---
train_ds = tf.keras.utils.image_dataset_from_directory(
    train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)
val_ds = tf.keras.utils.image_dataset_from_directory(
    val_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)
test_ds = tf.keras.utils.image_dataset_from_directory(
    test_dir, shuffle=False, batch_size=BATCH_SIZE, image_size=IMG_SIZE)

class_names = train_ds.class_names
NUM_CLASSES = len(class_names)
print(f"Classes ({NUM_CLASSES}): {class_names}")

# --- Helper: one-hot ---
def to_one_hot(image, label):
    return image, tf.one_hot(label, NUM_CLASSES)

# --- Beta distribution for MixUp/CutMix ---
def sample_beta_distribution(size, concentration_0=0.2, concentration_1=0.2):
    gamma_1 = tf.random.gamma(shape=[size], alpha=concentration_1, dtype=tf.float32)
    gamma_2 = tf.random.gamma(shape=[size], alpha=concentration_0, dtype=tf.float32)
    return gamma_2 / (gamma_1 + gamma_2 + 1e-8)

# --- MixUp ---
def mix_up(images, labels, alpha=0.2):
    batch_size = tf.shape(images)[0]
    lambda_param = sample_beta_distribution(batch_size, alpha, alpha)
    lambda_param_img = tf.reshape(lambda_param, [batch_size, 1, 1, 1])
    random_indices = tf.random.shuffle(tf.range(batch_size))
    mixed_images = lambda_param_img * images + (1 - lambda_param_img) * tf.gather(images, random_indices)
    labels = tf.cast(labels, tf.float32)
    lambda_param_lbl = tf.reshape(lambda_param, [-1, 1])
    mixed_labels = lambda_param_lbl * labels + (1 - lambda_param_lbl) * tf.gather(labels, random_indices)
    return mixed_images, mixed_labels

# --- CutMix (batch-level via coordinate-grid broadcasting) ---
def cut_mix(images, labels, alpha=0.2):
    batch_size = tf.shape(images)[0]
    h = tf.shape(images)[1]
    w = tf.shape(images)[2]
    
    lambda_param = sample_beta_distribution(batch_size, alpha, alpha)  # [B]
    random_indices = tf.random.shuffle(tf.range(batch_size))
    
    # Per-sample cut sizes (in pixels)
    cut_ratio = tf.sqrt(1.0 - lambda_param)  # [B]
    r_h = tf.cast(cut_ratio * tf.cast(h, tf.float32), tf.int32)  # [B]
    r_w = tf.cast(cut_ratio * tf.cast(w, tf.float32), tf.int32)  # [B]
    
    # Per-sample random cut centres
    cx = tf.random.uniform([batch_size], 0, w, tf.int32)  # [B]
    cy = tf.random.uniform([batch_size], 0, h, tf.int32)  # [B]
    
    half_h = r_h // 2
    half_w = r_w // 2
    x1 = tf.clip_by_value(cx - half_w, 0, w)  # [B]
    x2 = tf.clip_by_value(cx + half_w, 0, w)  # [B]
    y1 = tf.clip_by_value(cy - half_h, 0, h)  # [B]
    y2 = tf.clip_by_value(cy + half_h, 0, h)  # [B]
    
    # Build boolean cut mask via coordinate broadcasting
    col_idx = tf.range(w, dtype=tf.int32)          # [W]
    row_idx = tf.range(h, dtype=tf.int32)          # [H]
    
    in_x = tf.logical_and(
        tf.reshape(col_idx, [1, 1, w]) >= tf.reshape(x1, [batch_size, 1, 1]),
        tf.reshape(col_idx, [1, 1, w]) <  tf.reshape(x2, [batch_size, 1, 1])
    )                                               # [B, 1, W]
    in_y = tf.logical_and(
        tf.reshape(row_idx, [1, h, 1]) >= tf.reshape(y1, [batch_size, 1, 1]),
        tf.reshape(row_idx, [1, h, 1]) <  tf.reshape(y2, [batch_size, 1, 1])
    )                                               # [B, H, 1]
    
    cut_mask = tf.cast(tf.logical_and(in_y, in_x), tf.float32)  # [B, H, W]
    cut_mask = tf.expand_dims(cut_mask, -1)                     # [B, H, W, 1]
    
    # Mix images
    shuffled = tf.gather(images, random_indices)
    mixed_images = (1.0 - cut_mask) * images + cut_mask * shuffled
    
    # Mix labels
    labels = tf.cast(labels, tf.float32)
    lambda_reshaped = tf.reshape(lambda_param, [-1, 1])
    mixed_labels = lambda_reshaped * labels + (1.0 - lambda_reshaped) * tf.gather(labels, random_indices)
    
    return mixed_images, mixed_labels

# --- RandomErasing ---
def random_erasing(images, probability=0.5, scale=(0.02, 0.33), ratio=(0.3, 3.3)):
    batch_size = tf.shape(images)[0]
    h = tf.shape(images)[1]
    w = tf.shape(images)[2]
    
    target_area = tf.random.uniform([], scale[0], scale[1]) * tf.cast(h * w, tf.float32)
    aspect_ratio = tf.random.uniform([], ratio[0], ratio[1])
    erasing_h = tf.cast(tf.math.sqrt(target_area / aspect_ratio), tf.int32)
    erasing_w = tf.cast(tf.math.sqrt(target_area * aspect_ratio), tf.int32)
    erasing_h = tf.clip_by_value(erasing_h, 1, h - 1)
    erasing_w = tf.clip_by_value(erasing_w, 1, w - 1)
    
    cx = tf.random.uniform([], 0, w - erasing_w, tf.int32)
    cy = tf.random.uniform([], 0, h - erasing_h, tf.int32)
    
    # Build erase mask via coordinate broadcasting (same pattern as CutMix)
    col_idx = tf.range(w, dtype=tf.int32)
    row_idx = tf.range(h, dtype=tf.int32)
    
    in_x = tf.logical_and(col_idx >= cx, col_idx < cx + erasing_w)        # [W]
    in_y = tf.logical_and(row_idx >= cy, row_idx < cy + erasing_h)        # [H]
    erase_mask = tf.cast(tf.expand_dims(in_y, 1) & tf.expand_dims(in_x, 0), tf.float32)  # [H, W]
    erase_mask = tf.expand_dims(erase_mask, 0)   # [1, H, W]
    erase_mask = tf.expand_dims(erase_mask, -1)  # [1, H, W, 1]
    
    noise = tf.random.uniform([batch_size, erasing_h, erasing_w, 3], 0.0, 255.0, dtype=tf.float32)
    
    # Pad the noise patch to full image size
    paddings = [[0, 0], [cy, h - (cy + erasing_h)], [cx, w - (cx + erasing_w)], [0, 0]]
    noise_padded = tf.pad(noise, paddings, constant_values=0.0)
    
    erased = images * (1.0 - erase_mask) + noise_padded * erase_mask
    return tf.cond(
        tf.random.uniform([]) < probability,
        lambda: erased,
        lambda: images
    )

# --- Geometric augmentation (kept for TTA compatibility) ---
data_augmentation = tf.keras.Sequential([
    layers.RandomFlip("horizontal_and_vertical"),
    layers.RandomRotation(0.2),
    layers.RandomZoom(0.2),
    layers.RandomTranslation(0.1, 0.1),
    layers.RandomContrast(0.2),
    layers.RandomBrightness(0.2)
], name="data_augmentation")

# --- Augment + MixUp/CutMix + RandomErasing pipeline ---
def augment_and_mix(images, labels):
    # images are uint8 [0,255], labels are int
    # 1. Geometric augmentation
    images = data_augmentation(images, training=True)
    # 2. Convert to float32 for MixUp/CutMix
    images = tf.cast(images, tf.float32)
    # 3. Apply CutMix or MixUp with 50% probability
    choice = tf.random.uniform([])
    labels_onehot = tf.one_hot(labels, NUM_CLASSES)
    images, labels_onehot = tf.cond(
        choice < 0.3,  # 30% MixUp
        lambda: mix_up(images, labels_onehot),
        lambda: tf.cond(
            choice < 0.6,  # 30% CutMix
            lambda: cut_mix(images, labels_onehot),
            lambda: (images, labels_onehot)  # 40% no mix
        )
    )
    # 4. Random Erasing
    images = random_erasing(images, probability=0.25)
    return images, labels_onehot

# --- Preprocessing (EfficientNetV2: [-1,1]) ---
def preprocess_fn(image, label):
    return preprocess_input(image), label

# --- Class Weights ---
train_class_counts = Counter()
for cn in class_names:
    p = os.path.join(train_dir, cn)
    if os.path.exists(p):
        train_class_counts[cn] = len(os.listdir(p))

y_integer = []
for i, cn in enumerate(class_names):
    y_integer.extend([i] * train_class_counts[cn])

class_weights_array = compute_class_weight('balanced', classes=np.unique(y_integer), y=y_integer)
class_weights_capped = [min(w, 2.5) for w in class_weights_array]
class_weights_tensor = tf.constant(class_weights_capped, dtype=tf.float32)

def add_sample_weight(image, label):
    class_indices = tf.argmax(label, axis=-1)
    sample_weights = tf.gather(class_weights_tensor, class_indices)
    return image, label, sample_weights

print("Class weights (capped at 2.5):")
for i, cn in enumerate(class_names):
    print(f"  {cn}: {class_weights_capped[i]:.4f}")

# --- Build pipelines ---
# Train: augment → mix → erase → preprocess → onehot → sample_weight → prefetch
train_ds = (train_ds
    .map(augment_and_mix, num_parallel_calls=AUTOTUNE)
    .map(preprocess_fn, num_parallel_calls=AUTOTUNE)
    .map(add_sample_weight, num_parallel_calls=AUTOTUNE)
    .prefetch(AUTOTUNE))

# Val/Test: preprocess → onehot → prefetch
val_ds = (val_ds
    .map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)
    .map(to_one_hot, num_parallel_calls=AUTOTUNE)
    .prefetch(AUTOTUNE))

test_ds = (test_ds
    .map(lambda img, lbl: (preprocess_input(tf.cast(img, tf.float32)), lbl), num_parallel_calls=AUTOTUNE)
    .map(to_one_hot, num_parallel_calls=AUTOTUNE)
    .prefetch(AUTOTUNE))

print("Data pipelines ready.")

Found 4890 files belonging to 4 classes.
Found 1046 files belonging to 4 classes.
Found 1052 files belonging to 4 classes.
Classes (4): ['Bercak Daun', 'Daun Sehat', 'Hawar Daun', 'Karat Daun']
Class weights (capped at 2.5):
  Bercak Daun: 1.1665
  Daun Sehat: 0.8080
  Hawar Daun: 1.0847
  Karat Daun: 1.0171
Data pipelines ready.

6. Visualisasi Sampel Data (dengan Augmentasi)

In [18]:

# Visualisasi augmentasi (non-preprocessed)
vis_ds = tf.keras.utils.image_dataset_from_directory(
    train_dir, shuffle=True, batch_size=BATCH_SIZE, image_size=IMG_SIZE)

plt.figure(figsize=(16, 10))
for images, labels in vis_ds.take(1):
    # Original (4)
    for i in range(4):
        plt.subplot(3, 4, i + 1)
        plt.imshow(images[i].numpy().astype("uint8"))
        plt.title(f"Asli: {class_names[labels[i].numpy()]}", fontsize=11)
        plt.axis("off")
    # Augmented (4) — geometric only
    aug = data_augmentation(images, training=True)
    for i in range(4):
        plt.subplot(3, 4, i + 5)
        plt.imshow(aug[i].numpy().astype("uint8"))
        plt.title(f"Aug: {class_names[labels[i].numpy()]}", fontsize=11)
        plt.axis("off")
    # MixUp/CutMix result (4)
    aug_f = tf.cast(images, tf.float32)
    mixed, _ = mix_up(aug_f, tf.one_hot(labels, NUM_CLASSES))
    for i in range(4):
        plt.subplot(3, 4, i + 9)
        plt.imshow(tf.clip_by_value(mixed[i], 0, 255).numpy().astype("uint8"))
        plt.title("MixUp/CutMix", fontsize=11)
        plt.axis("off")

plt.suptitle("Contoh Augmentasi Berlapis", fontsize=16)
plt.tight_layout()
plt.show()

Found 4890 files belonging to 4 classes.

2026-06-11 17:05:59.854775: I tensorflow/core/framework/local_rendezvous.cc:407] Local rendezvous is aborting with status: OUT_OF_RANGE: End of sequence

7. Arsitektur Model CNN (EfficientNetV2B0)

Base model EfficientNetV2B0 (imagenet, frozen) + Conv2D tambahan + BatchNorm + Dropout + Dense. Ditambah GaussianNoise(0.1) untuk regularisasi tambahan.

In [19]:

def build_model(num_classes):
    base_model = EfficientNetV2B0(
        input_shape=IMG_SIZE + (3,),
        include_top=False,
        weights='imagenet',
    )
    base_model.trainable = False

    inputs = tf.keras.Input(shape=IMG_SIZE + (3,))
    x = layers.GaussianNoise(0.1)(inputs)
    x = base_model(x, training=False)
    x = layers.Conv2D(512, (3, 3), padding='same', activation='swish')(x)
    x = layers.BatchNormalization()(x)
    x = layers.MaxPooling2D((2, 2))(x)
    x = layers.Dropout(0.2)(x)
    x = layers.Conv2D(256, (3, 3), padding='same', activation='swish')(x)
    x = layers.BatchNormalization()(x)
    x = layers.GlobalAveragePooling2D()(x)
    x = layers.Dropout(0.3)(x)
    x = layers.Dense(1024, activation='swish')(x)
    x = layers.BatchNormalization()(x)
    x = layers.Dropout(0.4)(x)
    outputs = layers.Dense(num_classes, activation='softmax', dtype='float32')(x)
    return models.Model(inputs, outputs), base_model

if IS_COLAB:
    ckpt_dir = '/content/best_model'
else:
    ckpt_dir = os.path.join(os.getcwd(), 'best_model')
checkpoint_path = os.path.join(ckpt_dir, 'best_model.keras')

if os.path.exists(checkpoint_path):
    print(f"Loading checkpoint: {checkpoint_path}")
    try:
        model = models.load_model(checkpoint_path, compile=False)
    except Exception as e:
        print(f"Load failed: {e}. Building fresh.")
        model, base_model = build_model(NUM_CLASSES)
else:
    print("No checkpoint. Building fresh model.")
    model, base_model = build_model(NUM_CLASSES)

os.makedirs(ckpt_dir, exist_ok=True)
model.summary()

Loading checkpoint: /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Model: "functional_1"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ gaussian_noise (GaussianNoise)  │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ efficientnetv2-b0 (Functional)  │ (None, 7, 7, 1280)     │     5,919,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ conv2d (Conv2D)                 │ (None, 7, 7, 512)      │     5,898,752 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ batch_normalization             │ (None, 7, 7, 512)      │         2,048 │
│ (BatchNormalization)            │                        │               │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ max_pooling2d (MaxPooling2D)    │ (None, 3, 3, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout (Dropout)               │ (None, 3, 3, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ conv2d_1 (Conv2D)               │ (None, 3, 3, 256)      │     1,179,904 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ batch_normalization_1           │ (None, 3, 3, 256)      │         1,024 │
│ (BatchNormalization)            │                        │               │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ global_average_pooling2d        │ (None, 256)            │             0 │
│ (GlobalAveragePooling2D)        │                        │               │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout_1 (Dropout)             │ (None, 256)            │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense (Dense)                   │ (None, 1024)           │       263,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ batch_normalization_2           │ (None, 1024)           │         4,096 │
│ (BatchNormalization)            │                        │               │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dropout_2 (Dropout)             │ (None, 1024)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 4)              │         4,100 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 13,272,404 (50.63 MB)

 Trainable params: 7,349,508 (28.04 MB)

 Non-trainable params: 5,922,896 (22.59 MB)

8. Callbacks

ModelCheckpoint: simpan yang terbaik (val_accuracy)
EarlyStopping: patience 15 (lebih panjang untuk fine-tuning)
ReduceLROnPlateau: faktor 0.2, patience 5, min 1e-8
CSVLogger: riwayat training

In [20]:

import time

log_dir = "logs/fit/" + time.strftime("%Y%m%d-%H%M%S")

checkpoint_cb = callbacks.ModelCheckpoint(
    checkpoint_path,
    save_best_only=True,
    monitor="val_accuracy",
    mode="max",
    verbose=1
)

early_stopping_cb = callbacks.EarlyStopping(
    monitor="val_accuracy",
    patience=15,
    restore_best_weights=True,
    mode="max",
    verbose=1
)

reduce_lr_cb = callbacks.ReduceLROnPlateau(
    monitor='val_loss',
    factor=0.2,
    patience=5,
    min_lr=1e-8,
    verbose=1,
    mode="min"
)

csv_logger = callbacks.CSVLogger(os.path.join(ckpt_dir, 'training_log.csv'))

callbacks_list = [checkpoint_cb, early_stopping_cb, reduce_lr_cb, csv_logger]

9. Pelatihan Fase 1 — Head Only

Optimizer AdamW dengan EMA (Exponential Moving Average) dan Label Smoothing 0.2. Base model beku, hanya head yang dilatih.

In [21]:

model.compile(
    optimizer=AdamW(use_ema=True, ema_momentum=0.999,
                    learning_rate=1e-3, weight_decay=1e-4),
    loss=tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.2),
    metrics=['accuracy']
)

print("Fase 1: Head training (base frozen)...")
history_1 = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=30,
    callbacks=callbacks_list
)

Fase 1: Head training (base frozen)...
Epoch 1/30

Corrupt JPEG data: bad Huffman code

[1m 72/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m2s[0m 36ms/step - accuracy: 0.8908 - loss: 0.8013

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 135ms/step - accuracy: 0.8922 - loss: 0.8094
Epoch 1: val_accuracy improved from None to 0.95411, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 1: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m61s[0m 215ms/step - accuracy: 0.8941 - loss: 0.8123 - val_accuracy: 0.9541 - val_loss: 0.7203 - learning_rate: 0.0010
Epoch 2/30
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 68ms/step - accuracy: 0.9128 - loss: 0.7974

Corrupt JPEG data: bad Huffman code

[1m 64/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 72ms/step - accuracy: 0.9082 - loss: 0.8043

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 69ms/step - accuracy: 0.9014 - loss: 0.8124
Epoch 2: val_accuracy improved from 0.95411 to 0.95985, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 2: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 94ms/step - accuracy: 0.8957 - loss: 0.8151 - val_accuracy: 0.9598 - val_loss: 0.7770 - learning_rate: 0.0010
Epoch 3/30
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 68ms/step - accuracy: 0.8883 - loss: 0.8339

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m5s[0m 68ms/step - accuracy: 0.8923 - loss: 0.8253

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 68ms/step - accuracy: 0.8969 - loss: 0.8170
Epoch 3: val_accuracy improved from 0.95985 to 0.97514, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 3: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 96ms/step - accuracy: 0.9033 - loss: 0.8026 - val_accuracy: 0.9751 - val_loss: 0.6700 - learning_rate: 0.0010
Epoch 4/30
[1m 38/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 78ms/step - accuracy: 0.8932 - loss: 0.8082

Corrupt JPEG data: bad Huffman code

[1m 72/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 72ms/step - accuracy: 0.8967 - loss: 0.8057

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.8977 - loss: 0.8050
Epoch 4: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 84ms/step - accuracy: 0.8984 - loss: 0.8053 - val_accuracy: 0.9618 - val_loss: 0.6673 - learning_rate: 0.0010
Epoch 5/30
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 82ms/step - accuracy: 0.8997 - loss: 0.8094

Corrupt JPEG data: bad Huffman code

[1m 87/153[0m [32m━━━━━━━━━━━[0m[37m━━━━━━━━━[0m [1m4s[0m 71ms/step - accuracy: 0.8984 - loss: 0.8092

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 68ms/step - accuracy: 0.9004 - loss: 0.8062
Epoch 5: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 82ms/step - accuracy: 0.9055 - loss: 0.7987 - val_accuracy: 0.9512 - val_loss: 0.7019 - learning_rate: 0.0010
Epoch 6/30
[1m 40/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 77ms/step - accuracy: 0.9069 - loss: 0.7904

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 71ms/step - accuracy: 0.9013 - loss: 0.8002

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.8983 - loss: 0.8076
Epoch 6: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 82ms/step - accuracy: 0.8965 - loss: 0.8128 - val_accuracy: 0.9646 - val_loss: 0.6677 - learning_rate: 0.0010
Epoch 7/30
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 82ms/step - accuracy: 0.9011 - loss: 0.7930

Corrupt JPEG data: bad Huffman code

[1m 76/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 75ms/step - accuracy: 0.8995 - loss: 0.8002

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 71ms/step - accuracy: 0.9031 - loss: 0.7991
Epoch 7: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 84ms/step - accuracy: 0.9065 - loss: 0.7975 - val_accuracy: 0.9685 - val_loss: 0.6607 - learning_rate: 0.0010
Epoch 8/30
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 74ms/step - accuracy: 0.9106 - loss: 0.7855

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 71ms/step - accuracy: 0.9116 - loss: 0.7886

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 69ms/step - accuracy: 0.9099 - loss: 0.7915
Epoch 8: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 84ms/step - accuracy: 0.9108 - loss: 0.7886 - val_accuracy: 0.9732 - val_loss: 0.6645 - learning_rate: 0.0010
Epoch 9/30
[1m 38/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 64ms/step - accuracy: 0.9102 - loss: 0.7870

Corrupt JPEG data: bad Huffman code

[1m 74/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 75ms/step - accuracy: 0.9106 - loss: 0.7890

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9121 - loss: 0.7885
Epoch 9: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9119 - loss: 0.7855 - val_accuracy: 0.9474 - val_loss: 0.6715 - learning_rate: 0.0010
Epoch 10/30
[1m 41/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 72ms/step - accuracy: 0.9041 - loss: 0.7846

Corrupt JPEG data: bad Huffman code

[1m 84/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m4s[0m 70ms/step - accuracy: 0.9028 - loss: 0.7952

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 69ms/step - accuracy: 0.9047 - loss: 0.7963
Epoch 10: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 82ms/step - accuracy: 0.9102 - loss: 0.7911 - val_accuracy: 0.9656 - val_loss: 0.6569 - learning_rate: 0.0010
Epoch 11/30
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 66ms/step - accuracy: 0.9093 - loss: 0.7968

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 69ms/step - accuracy: 0.9116 - loss: 0.7899

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 68ms/step - accuracy: 0.9124 - loss: 0.7892
Epoch 11: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m20s[0m 81ms/step - accuracy: 0.9110 - loss: 0.7865 - val_accuracy: 0.9522 - val_loss: 0.6961 - learning_rate: 0.0010
Epoch 12/30
[1m 43/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m7s[0m 66ms/step - accuracy: 0.9099 - loss: 0.7873

Corrupt JPEG data: bad Huffman code

[1m 82/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m4s[0m 68ms/step - accuracy: 0.9096 - loss: 0.7893

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 67ms/step - accuracy: 0.9108 - loss: 0.7897
Epoch 12: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 80ms/step - accuracy: 0.9147 - loss: 0.7860 - val_accuracy: 0.9627 - val_loss: 0.6555 - learning_rate: 0.0010
Epoch 13/30
[1m 38/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m6s[0m 60ms/step - accuracy: 0.8981 - loss: 0.7914

Corrupt JPEG data: bad Huffman code

[1m 66/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m5s[0m 68ms/step - accuracy: 0.9016 - loss: 0.7904

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 66ms/step - accuracy: 0.9061 - loss: 0.7873
Epoch 13: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 82ms/step - accuracy: 0.9129 - loss: 0.7795 - val_accuracy: 0.9560 - val_loss: 0.6718 - learning_rate: 0.0010
Epoch 14/30
[1m 56/153[0m [32m━━━━━━━[0m[37m━━━━━━━━━━━━━[0m [1m7s[0m 75ms/step - accuracy: 0.9193 - loss: 0.7872

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9188 - loss: 0.7865

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 74ms/step - accuracy: 0.9196 - loss: 0.7812
Epoch 14: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 89ms/step - accuracy: 0.9180 - loss: 0.7753 - val_accuracy: 0.9570 - val_loss: 0.6654 - learning_rate: 0.0010
Epoch 15/30
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 74ms/step - accuracy: 0.9333 - loss: 0.7673

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m5s[0m 69ms/step - accuracy: 0.9285 - loss: 0.7714

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 69ms/step - accuracy: 0.9224 - loss: 0.7744
Epoch 15: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 83ms/step - accuracy: 0.9172 - loss: 0.7775 - val_accuracy: 0.9589 - val_loss: 0.6675 - learning_rate: 0.0010
Epoch 16/30
[1m 29/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m9s[0m 73ms/step - accuracy: 0.9020 - loss: 0.7942

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 72ms/step - accuracy: 0.9073 - loss: 0.7869

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9098 - loss: 0.7838
Epoch 16: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9135 - loss: 0.7756 - val_accuracy: 0.9656 - val_loss: 0.6626 - learning_rate: 0.0010
Epoch 17/30
[1m 36/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 78ms/step - accuracy: 0.9200 - loss: 0.7642

Corrupt JPEG data: bad Huffman code

[1m 67/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 78ms/step - accuracy: 0.9167 - loss: 0.7716

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9165 - loss: 0.7747
Epoch 17: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9160 - loss: 0.7748 - val_accuracy: 0.9723 - val_loss: 0.6553 - learning_rate: 0.0010
Epoch 18/30
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 77ms/step - accuracy: 0.9418 - loss: 0.7480

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 78ms/step - accuracy: 0.9343 - loss: 0.7633

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9280 - loss: 0.7724
Epoch 18: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9223 - loss: 0.7801 - val_accuracy: 0.9713 - val_loss: 0.6571 - learning_rate: 0.0010
Epoch 18: early stopping
Restoring model weights from the end of the best epoch: 3.

10. Fine-tuning Fase 2 — Unfreeze Layer Atas

Membuka 100 layer teratas base model. Learning rate diturunkan ke 1e-4.

In [22]:

base_model = model.layers[2]  # layer[0]=InputLayer, layer[1]=GaussianNoise, layer[2]=EfficientNetV2B0
base_model.trainable = True
for layer in base_model.layers[:-100]:
    layer.trainable = False

model.compile(
    optimizer=AdamW(use_ema=True, ema_momentum=0.999,
                    learning_rate=1e-4, weight_decay=1e-4),
    loss=tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.2),
    metrics=['accuracy']
)

print("Fase 2: Fine-tuning top 100 layers...")
history_2 = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=60,
    initial_epoch=history_1.epoch[-1] + 1,
    callbacks=callbacks_list
)

Fase 2: Fine-tuning top 100 layers...
Epoch 19/60

Corrupt JPEG data: bad Huffman code
2026-06-11 17:11:41.079500: 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.
2026-06-11 17:11:41.237435: 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.

[1m 78/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m2s[0m 39ms/step - accuracy: 0.8603 - loss: 0.8519

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m150/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 51ms/step - accuracy: 0.8652 - loss: 0.8433

2026-06-11 17:12:10.589920: 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.
2026-06-11 17:12:10.748299: 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.

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 194ms/step - accuracy: 0.8654 - loss: 0.8430
Epoch 19: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m89s[0m 259ms/step - accuracy: 0.8763 - loss: 0.8239 - val_accuracy: 0.9637 - val_loss: 0.6448 - learning_rate: 1.0000e-04
Epoch 20/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 67ms/step - accuracy: 0.8962 - loss: 0.8058

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 70ms/step - accuracy: 0.9015 - loss: 0.8021

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 67ms/step - accuracy: 0.9021 - loss: 0.7997
Epoch 20: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 82ms/step - accuracy: 0.9033 - loss: 0.7931 - val_accuracy: 0.9656 - val_loss: 0.6464 - learning_rate: 1.0000e-04
Epoch 21/60
[1m 38/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 80ms/step - accuracy: 0.9096 - loss: 0.7774

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 75ms/step - accuracy: 0.9055 - loss: 0.7842

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 70ms/step - accuracy: 0.9021 - loss: 0.7901
Epoch 21: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9027 - loss: 0.7900 - val_accuracy: 0.9637 - val_loss: 0.6389 - learning_rate: 1.0000e-04
Epoch 22/60
[1m 41/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m9s[0m 86ms/step - accuracy: 0.9098 - loss: 0.7861

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9108 - loss: 0.7885

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 74ms/step - accuracy: 0.9111 - loss: 0.7886
Epoch 22: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 89ms/step - accuracy: 0.9096 - loss: 0.7858 - val_accuracy: 0.9751 - val_loss: 0.6314 - learning_rate: 1.0000e-04
Epoch 23/60
[1m 46/153[0m [32m━━━━━━[0m[37m━━━━━━━━━━━━━━[0m [1m8s[0m 77ms/step - accuracy: 0.9281 - loss: 0.7582

Corrupt JPEG data: bad Huffman code

[1m 72/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9233 - loss: 0.7659

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 73ms/step - accuracy: 0.9193 - loss: 0.7723
Epoch 23: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 88ms/step - accuracy: 0.9176 - loss: 0.7735 - val_accuracy: 0.9751 - val_loss: 0.6300 - learning_rate: 1.0000e-04
Epoch 24/60
[1m 30/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m9s[0m 81ms/step - accuracy: 0.9006 - loss: 0.7649

Corrupt JPEG data: bad Huffman code

[1m 76/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 72ms/step - accuracy: 0.9058 - loss: 0.7704

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9101 - loss: 0.7699
Epoch 24: val_accuracy did not improve from 0.97514
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9172 - loss: 0.7660 - val_accuracy: 0.9723 - val_loss: 0.6350 - learning_rate: 1.0000e-04
Epoch 25/60
[1m 50/153[0m [32m━━━━━━[0m[37m━━━━━━━━━━━━━━[0m [1m7s[0m 77ms/step - accuracy: 0.9242 - loss: 0.7608

Corrupt JPEG data: bad Huffman code

[1m 72/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9237 - loss: 0.7632

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 71ms/step - accuracy: 0.9236 - loss: 0.7635
Epoch 25: val_accuracy improved from 0.97514 to 0.97897, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 25: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m16s[0m 98ms/step - accuracy: 0.9239 - loss: 0.7602 - val_accuracy: 0.9790 - val_loss: 0.6311 - learning_rate: 1.0000e-04
Epoch 26/60
[1m 38/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 67ms/step - accuracy: 0.9397 - loss: 0.7498

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m5s[0m 70ms/step - accuracy: 0.9341 - loss: 0.7563

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9308 - loss: 0.7591
Epoch 26: val_accuracy did not improve from 0.97897
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9284 - loss: 0.7577 - val_accuracy: 0.9780 - val_loss: 0.6284 - learning_rate: 1.0000e-04
Epoch 27/60
[1m 30/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m10s[0m 83ms/step - accuracy: 0.9301 - loss: 0.7455

Corrupt JPEG data: bad Huffman code

[1m 84/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 76ms/step - accuracy: 0.9292 - loss: 0.7529

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9306 - loss: 0.7538
Epoch 27: val_accuracy did not improve from 0.97897
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9339 - loss: 0.7496 - val_accuracy: 0.9780 - val_loss: 0.6246 - learning_rate: 1.0000e-04
Epoch 28/60
[1m 64/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 74ms/step - accuracy: 0.9320 - loss: 0.7515

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2
Corrupt JPEG data: bad Huffman code

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 69ms/step - accuracy: 0.9291 - loss: 0.7553
Epoch 28: val_accuracy did not improve from 0.97897
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9278 - loss: 0.7560 - val_accuracy: 0.9771 - val_loss: 0.6247 - learning_rate: 1.0000e-04
Epoch 29/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m10s[0m 83ms/step - accuracy: 0.9392 - loss: 0.7419

Corrupt JPEG data: bad Huffman code

[1m 65/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 79ms/step - accuracy: 0.9357 - loss: 0.7439

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9338 - loss: 0.7478
Epoch 29: val_accuracy improved from 0.97897 to 0.98088, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 29: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 95ms/step - accuracy: 0.9321 - loss: 0.7475 - val_accuracy: 0.9809 - val_loss: 0.6241 - learning_rate: 1.0000e-04
Epoch 30/60
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 69ms/step - accuracy: 0.9211 - loss: 0.7697

Corrupt JPEG data: bad Huffman code

[1m 66/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 72ms/step - accuracy: 0.9290 - loss: 0.7616

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 68ms/step - accuracy: 0.9309 - loss: 0.7562
Epoch 30: val_accuracy improved from 0.98088 to 0.98279, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 30: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 90ms/step - accuracy: 0.9319 - loss: 0.7498 - val_accuracy: 0.9828 - val_loss: 0.6244 - learning_rate: 1.0000e-04
Epoch 31/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m10s[0m 85ms/step - accuracy: 0.9480 - loss: 0.7287

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9436 - loss: 0.7344

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9419 - loss: 0.7386
Epoch 31: val_accuracy did not improve from 0.98279
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9393 - loss: 0.7419 - val_accuracy: 0.9828 - val_loss: 0.6205 - learning_rate: 1.0000e-04
Epoch 32/60
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 75ms/step - accuracy: 0.9310 - loss: 0.7688

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 74ms/step - accuracy: 0.9300 - loss: 0.7629

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 71ms/step - accuracy: 0.9298 - loss: 0.7579
Epoch 32: val_accuracy did not improve from 0.98279
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9313 - loss: 0.7486 - val_accuracy: 0.9828 - val_loss: 0.6224 - learning_rate: 1.0000e-04
Epoch 33/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 71ms/step - accuracy: 0.9487 - loss: 0.7287

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9475 - loss: 0.7323

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9430 - loss: 0.7388
Epoch 33: val_accuracy did not improve from 0.98279
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 90ms/step - accuracy: 0.9374 - loss: 0.7429 - val_accuracy: 0.9809 - val_loss: 0.6203 - learning_rate: 1.0000e-04
Epoch 34/60
[1m 36/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 80ms/step - accuracy: 0.9398 - loss: 0.7303

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 75ms/step - accuracy: 0.9384 - loss: 0.7403

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9367 - loss: 0.7442
Epoch 34: val_accuracy did not improve from 0.98279
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9374 - loss: 0.7437 - val_accuracy: 0.9790 - val_loss: 0.6263 - learning_rate: 1.0000e-04
Epoch 35/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 79ms/step - accuracy: 0.9413 - loss: 0.7367

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9405 - loss: 0.7398

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 74ms/step - accuracy: 0.9404 - loss: 0.7415
Epoch 35: val_accuracy did not improve from 0.98279
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 89ms/step - accuracy: 0.9395 - loss: 0.7400 - val_accuracy: 0.9828 - val_loss: 0.6210 - learning_rate: 1.0000e-04
Epoch 36/60
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 77ms/step - accuracy: 0.9411 - loss: 0.7312

Corrupt JPEG data: bad Huffman code

[1m 65/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9378 - loss: 0.7412

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9379 - loss: 0.7430
Epoch 36: val_accuracy improved from 0.98279 to 0.98566, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 36: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 93ms/step - accuracy: 0.9378 - loss: 0.7419 - val_accuracy: 0.9857 - val_loss: 0.6177 - learning_rate: 1.0000e-04
Epoch 37/60
[1m 39/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 72ms/step - accuracy: 0.9443 - loss: 0.7402

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 71ms/step - accuracy: 0.9412 - loss: 0.7417

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 68ms/step - accuracy: 0.9393 - loss: 0.7431
Epoch 37: val_accuracy did not improve from 0.98566
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 82ms/step - accuracy: 0.9389 - loss: 0.7403 - val_accuracy: 0.9847 - val_loss: 0.6154 - learning_rate: 1.0000e-04
Epoch 38/60
[1m 37/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 82ms/step - accuracy: 0.9461 - loss: 0.7304

Corrupt JPEG data: bad Huffman code

[1m117/153[0m [32m━━━━━━━━━━━━━━━[0m[37m━━━━━[0m [1m2s[0m 73ms/step - accuracy: 0.9463 - loss: 0.7330

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9453 - loss: 0.7337
Epoch 38: val_accuracy improved from 0.98566 to 0.98662, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 38: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 96ms/step - accuracy: 0.9417 - loss: 0.7345 - val_accuracy: 0.9866 - val_loss: 0.6151 - learning_rate: 1.0000e-04
Epoch 39/60
[1m 42/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 79ms/step - accuracy: 0.9449 - loss: 0.7251

Corrupt JPEG data: bad Huffman code

[1m 71/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 79ms/step - accuracy: 0.9439 - loss: 0.7302

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 73ms/step - accuracy: 0.9420 - loss: 0.7343
Epoch 39: val_accuracy did not improve from 0.98662
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9397 - loss: 0.7370 - val_accuracy: 0.9818 - val_loss: 0.6191 - learning_rate: 1.0000e-04
Epoch 40/60
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 79ms/step - accuracy: 0.9374 - loss: 0.7308

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 75ms/step - accuracy: 0.9419 - loss: 0.7314

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 71ms/step - accuracy: 0.9420 - loss: 0.7341
Epoch 40: val_accuracy improved from 0.98662 to 0.98757, saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras

Epoch 40: finished saving model to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m15s[0m 96ms/step - accuracy: 0.9440 - loss: 0.7332 - val_accuracy: 0.9876 - val_loss: 0.6149 - learning_rate: 1.0000e-04
Epoch 41/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 67ms/step - accuracy: 0.9416 - loss: 0.7311

Corrupt JPEG data: bad Huffman code

[1m 77/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 77ms/step - accuracy: 0.9450 - loss: 0.7311

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 70ms/step - accuracy: 0.9461 - loss: 0.7309
Epoch 41: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9476 - loss: 0.7281 - val_accuracy: 0.9828 - val_loss: 0.6191 - learning_rate: 1.0000e-04
Epoch 42/60
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 82ms/step - accuracy: 0.9547 - loss: 0.7103

Corrupt JPEG data: bad Huffman code

[1m 76/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 75ms/step - accuracy: 0.9456 - loss: 0.7218

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9441 - loss: 0.7269
Epoch 42: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 83ms/step - accuracy: 0.9464 - loss: 0.7277 - val_accuracy: 0.9857 - val_loss: 0.6179 - learning_rate: 1.0000e-04
Epoch 43/60
[1m 30/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m10s[0m 84ms/step - accuracy: 0.9397 - loss: 0.7041

Corrupt JPEG data: bad Huffman code

[1m 67/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 75ms/step - accuracy: 0.9422 - loss: 0.7177

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9454 - loss: 0.7238
Epoch 43: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9501 - loss: 0.7242 - val_accuracy: 0.9857 - val_loss: 0.6156 - learning_rate: 1.0000e-04
Epoch 44/60
[1m 37/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 86ms/step - accuracy: 0.9436 - loss: 0.7309

Corrupt JPEG data: bad Huffman code

[1m 80/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 79ms/step - accuracy: 0.9452 - loss: 0.7302

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9444 - loss: 0.7302
Epoch 44: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9436 - loss: 0.7278 - val_accuracy: 0.9866 - val_loss: 0.6154 - learning_rate: 1.0000e-04
Epoch 45/60
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m10s[0m 89ms/step - accuracy: 0.9404 - loss: 0.7367

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 80ms/step - accuracy: 0.9442 - loss: 0.7334

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 74ms/step - accuracy: 0.9477 - loss: 0.7307
Epoch 45: val_accuracy did not improve from 0.98757

Epoch 45: ReduceLROnPlateau reducing learning rate to 1.9999999494757503e-05.
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9501 - loss: 0.7272 - val_accuracy: 0.9828 - val_loss: 0.6205 - learning_rate: 1.0000e-04
Epoch 46/60
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 66ms/step - accuracy: 0.9486 - loss: 0.7205

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m5s[0m 66ms/step - accuracy: 0.9491 - loss: 0.7207

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 65ms/step - accuracy: 0.9501 - loss: 0.7230
Epoch 46: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 79ms/step - accuracy: 0.9507 - loss: 0.7220 - val_accuracy: 0.9866 - val_loss: 0.6155 - learning_rate: 2.0000e-05
Epoch 47/60
[1m 40/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 75ms/step - accuracy: 0.9507 - loss: 0.7353

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9503 - loss: 0.7324

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9493 - loss: 0.7299
Epoch 47: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9470 - loss: 0.7256 - val_accuracy: 0.9857 - val_loss: 0.6153 - learning_rate: 2.0000e-05
Epoch 48/60
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 70ms/step - accuracy: 0.9311 - loss: 0.7394

Corrupt JPEG data: bad Huffman code

[1m 77/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 74ms/step - accuracy: 0.9393 - loss: 0.7374

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9450 - loss: 0.7312
Epoch 48: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 88ms/step - accuracy: 0.9528 - loss: 0.7190 - val_accuracy: 0.9857 - val_loss: 0.6142 - learning_rate: 2.0000e-05
Epoch 49/60
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 68ms/step - accuracy: 0.9430 - loss: 0.7231

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 71ms/step - accuracy: 0.9468 - loss: 0.7197

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9486 - loss: 0.7203
Epoch 49: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9495 - loss: 0.7192 - val_accuracy: 0.9857 - val_loss: 0.6142 - learning_rate: 2.0000e-05
Epoch 50/60
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 79ms/step - accuracy: 0.9544 - loss: 0.7062

Corrupt JPEG data: bad Huffman code

[1m 77/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 77ms/step - accuracy: 0.9543 - loss: 0.7156

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9544 - loss: 0.7201
Epoch 50: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9556 - loss: 0.7218 - val_accuracy: 0.9857 - val_loss: 0.6141 - learning_rate: 2.0000e-05
Epoch 51/60
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 75ms/step - accuracy: 0.9369 - loss: 0.7202

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 71ms/step - accuracy: 0.9399 - loss: 0.7260

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 68ms/step - accuracy: 0.9426 - loss: 0.7270
Epoch 51: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 83ms/step - accuracy: 0.9485 - loss: 0.7215 - val_accuracy: 0.9866 - val_loss: 0.6120 - learning_rate: 2.0000e-05
Epoch 52/60
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 81ms/step - accuracy: 0.9490 - loss: 0.7133

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9513 - loss: 0.7147

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9511 - loss: 0.7186
Epoch 52: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 83ms/step - accuracy: 0.9515 - loss: 0.7192 - val_accuracy: 0.9866 - val_loss: 0.6140 - learning_rate: 2.0000e-05
Epoch 53/60
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 78ms/step - accuracy: 0.9581 - loss: 0.7043

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 77ms/step - accuracy: 0.9552 - loss: 0.7114

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9532 - loss: 0.7162
Epoch 53: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9519 - loss: 0.7191 - val_accuracy: 0.9866 - val_loss: 0.6126 - learning_rate: 2.0000e-05
Epoch 54/60
[1m 34/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m7s[0m 65ms/step - accuracy: 0.9428 - loss: 0.7223

Corrupt JPEG data: bad Huffman code

[1m 72/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 67ms/step - accuracy: 0.9470 - loss: 0.7227

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m151/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 68ms/step - accuracy: 0.9497 - loss: 0.7232
Epoch 54: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 83ms/step - accuracy: 0.9519 - loss: 0.7192 - val_accuracy: 0.9866 - val_loss: 0.6121 - learning_rate: 2.0000e-05
Epoch 55/60
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 81ms/step - accuracy: 0.9524 - loss: 0.7033

Corrupt JPEG data: bad Huffman code

[1m 65/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 78ms/step - accuracy: 0.9525 - loss: 0.7081

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9527 - loss: 0.7140
Epoch 55: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m20s[0m 84ms/step - accuracy: 0.9528 - loss: 0.7169 - val_accuracy: 0.9837 - val_loss: 0.6128 - learning_rate: 2.0000e-05
Epoch 55: early stopping
Restoring model weights from the end of the best epoch: 40.

11. Fine-tuning Fase 3 — Full Unfreeze

Membuka semua layer base model. Learning rate 5e-5 (sangat kecil agar tidak merusak bobot pretrained).

In [23]:

# Unfreeze all layers in base model
for layer in base_model.layers:
    layer.trainable = True

model.compile(
    optimizer=AdamW(use_ema=True, ema_momentum=0.999,
                    learning_rate=5e-5, weight_decay=1e-4),
    loss=tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.2),
    metrics=['accuracy']
)

print("Fase 3: Full fine-tuning...")
history_3 = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=90,
    initial_epoch=history_2.epoch[-1] + 1,
    callbacks=callbacks_list
)

Fase 3: Full fine-tuning...
Epoch 56/90

Corrupt JPEG data: bad Huffman code
2026-06-11 17:22:10.365783: 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.
2026-06-11 17:22:10.516772: 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.
2026-06-11 17:22:11.003935: 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.
2026-06-11 17:22:11.170998: 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.
2026-06-11 17:22:15.705420: 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.
2026-06-11 17:22:15.856916: 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.

[1m 48/153[0m [32m━━━━━━[0m[37m━━━━━━━━━━━━━━[0m [1m5s[0m 52ms/step - accuracy: 0.8750 - loss: 0.8173

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 55ms/step - accuracy: 0.8887 - loss: 0.8050

2026-06-11 17:23:02.858842: 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.
2026-06-11 17:23:03.009825: 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.
2026-06-11 17:23:03.495720: 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.
2026-06-11 17:23:03.662273: 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.
2026-06-11 17:23:07.775573: 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.
2026-06-11 17:23:07.927092: 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.

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 351ms/step - accuracy: 0.8887 - loss: 0.8049
Epoch 56: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m157s[0m 417ms/step - accuracy: 0.9000 - loss: 0.7896 - val_accuracy: 0.9646 - val_loss: 0.6448 - learning_rate: 5.0000e-05
Epoch 57/90
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 80ms/step - accuracy: 0.9124 - loss: 0.7628

Corrupt JPEG data: bad Huffman code

[1m 99/153[0m [32m━━━━━━━━━━━━[0m[37m━━━━━━━━[0m [1m4s[0m 75ms/step - accuracy: 0.9152 - loss: 0.7619

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9167 - loss: 0.7615
Epoch 57: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9213 - loss: 0.7588 - val_accuracy: 0.9732 - val_loss: 0.6351 - learning_rate: 5.0000e-05
Epoch 58/90
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 72ms/step - accuracy: 0.9371 - loss: 0.7455

Corrupt JPEG data: bad Huffman code

[1m 73/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 73ms/step - accuracy: 0.9333 - loss: 0.7503

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 67ms/step - accuracy: 0.9298 - loss: 0.7535
Epoch 58: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 82ms/step - accuracy: 0.9260 - loss: 0.7548 - val_accuracy: 0.9761 - val_loss: 0.6319 - learning_rate: 5.0000e-05
Epoch 59/90
[1m 33/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m10s[0m 87ms/step - accuracy: 0.9335 - loss: 0.7396

Corrupt JPEG data: bad Huffman code

[1m 75/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 78ms/step - accuracy: 0.9295 - loss: 0.7522

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 72ms/step - accuracy: 0.9292 - loss: 0.7565
Epoch 59: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m21s[0m 85ms/step - accuracy: 0.9321 - loss: 0.7544 - val_accuracy: 0.9771 - val_loss: 0.6291 - learning_rate: 5.0000e-05
Epoch 60/90
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 75ms/step - accuracy: 0.9324 - loss: 0.7550

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9322 - loss: 0.7533

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9339 - loss: 0.7503
Epoch 60: val_accuracy did not improve from 0.98757

Epoch 60: ReduceLROnPlateau reducing learning rate to 9.999999747378752e-06.
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9348 - loss: 0.7452 - val_accuracy: 0.9685 - val_loss: 0.6415 - learning_rate: 5.0000e-05
Epoch 61/90
[1m 29/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m10s[0m 89ms/step - accuracy: 0.9491 - loss: 0.7270

Corrupt JPEG data: bad Huffman code

[1m 70/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9471 - loss: 0.7332

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9448 - loss: 0.7365
Epoch 61: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 85ms/step - accuracy: 0.9393 - loss: 0.7366 - val_accuracy: 0.9751 - val_loss: 0.6266 - learning_rate: 1.0000e-05
Epoch 62/90
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 81ms/step - accuracy: 0.9465 - loss: 0.7358

Corrupt JPEG data: bad Huffman code

[1m 95/153[0m [32m━━━━━━━━━━━━[0m[37m━━━━━━━━[0m [1m4s[0m 77ms/step - accuracy: 0.9391 - loss: 0.7420

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9384 - loss: 0.7421
Epoch 62: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 90ms/step - accuracy: 0.9387 - loss: 0.7385 - val_accuracy: 0.9771 - val_loss: 0.6232 - learning_rate: 1.0000e-05
Epoch 63/90
[1m 47/153[0m [32m━━━━━━[0m[37m━━━━━━━━━━━━━━[0m [1m8s[0m 76ms/step - accuracy: 0.9410 - loss: 0.7559

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 76ms/step - accuracy: 0.9399 - loss: 0.7529

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 72ms/step - accuracy: 0.9388 - loss: 0.7501
Epoch 63: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9403 - loss: 0.7426 - val_accuracy: 0.9742 - val_loss: 0.6271 - learning_rate: 1.0000e-05
Epoch 64/90
[1m 35/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 84ms/step - accuracy: 0.9469 - loss: 0.7338

Corrupt JPEG data: bad Huffman code

[1m 76/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m5s[0m 77ms/step - accuracy: 0.9449 - loss: 0.7385

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 73ms/step - accuracy: 0.9452 - loss: 0.7382
Epoch 64: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9436 - loss: 0.7356 - val_accuracy: 0.9771 - val_loss: 0.6201 - learning_rate: 1.0000e-05
Epoch 65/90
[1m 29/153[0m [32m━━━[0m[37m━━━━━━━━━━━━━━━━━[0m [1m10s[0m 85ms/step - accuracy: 0.9236 - loss: 0.7620

Corrupt JPEG data: bad Huffman code

[1m 65/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 75ms/step - accuracy: 0.9298 - loss: 0.7583

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 69ms/step - accuracy: 0.9346 - loss: 0.7528
Epoch 65: val_accuracy did not improve from 0.98757

Epoch 65: ReduceLROnPlateau reducing learning rate to 1.9999999494757505e-06.
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m13s[0m 84ms/step - accuracy: 0.9407 - loss: 0.7423 - val_accuracy: 0.9761 - val_loss: 0.6246 - learning_rate: 1.0000e-05
Epoch 66/90
[1m 36/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m9s[0m 82ms/step - accuracy: 0.9442 - loss: 0.7218

Corrupt JPEG data: bad Huffman code

[1m 68/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 73ms/step - accuracy: 0.9456 - loss: 0.7259

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 69ms/step - accuracy: 0.9451 - loss: 0.7287
Epoch 66: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 84ms/step - accuracy: 0.9452 - loss: 0.7264 - val_accuracy: 0.9761 - val_loss: 0.6244 - learning_rate: 2.0000e-06
Epoch 67/90
[1m 32/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m8s[0m 70ms/step - accuracy: 0.9512 - loss: 0.7123

Corrupt JPEG data: bad Huffman code

[1m 65/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 70ms/step - accuracy: 0.9483 - loss: 0.7208

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 70ms/step - accuracy: 0.9436 - loss: 0.7282
Epoch 67: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 83ms/step - accuracy: 0.9407 - loss: 0.7299 - val_accuracy: 0.9780 - val_loss: 0.6249 - learning_rate: 2.0000e-06
Epoch 68/90
[1m 53/153[0m [32m━━━━━━[0m[37m━━━━━━━━━━━━━━[0m [1m7s[0m 74ms/step - accuracy: 0.9455 - loss: 0.7280

Corrupt JPEG data: bad Huffman code

[1m 67/153[0m [32m━━━━━━━━[0m[37m━━━━━━━━━━━━[0m [1m6s[0m 72ms/step - accuracy: 0.9457 - loss: 0.7291

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 70ms/step - accuracy: 0.9452 - loss: 0.7308
Epoch 68: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9442 - loss: 0.7296 - val_accuracy: 0.9780 - val_loss: 0.6240 - learning_rate: 2.0000e-06
Epoch 69/90
[1m 43/153[0m [32m━━━━━[0m[37m━━━━━━━━━━━━━━━[0m [1m8s[0m 77ms/step - accuracy: 0.9282 - loss: 0.7398

Corrupt JPEG data: bad Huffman code

[1m 82/153[0m [32m━━━━━━━━━━[0m[37m━━━━━━━━━━[0m [1m5s[0m 75ms/step - accuracy: 0.9330 - loss: 0.7367

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 72ms/step - accuracy: 0.9374 - loss: 0.7340
Epoch 69: val_accuracy did not improve from 0.98757
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 87ms/step - accuracy: 0.9423 - loss: 0.7319 - val_accuracy: 0.9751 - val_loss: 0.6239 - learning_rate: 2.0000e-06
Epoch 70/90
[1m 31/153[0m [32m━━━━[0m[37m━━━━━━━━━━━━━━━━[0m [1m10s[0m 86ms/step - accuracy: 0.9618 - loss: 0.7141

Corrupt JPEG data: bad Huffman code

[1m 69/153[0m [32m━━━━━━━━━[0m[37m━━━━━━━━━━━[0m [1m6s[0m 78ms/step - accuracy: 0.9542 - loss: 0.7217

Corrupt JPEG data: 75969 extraneous bytes before marker 0xd2

[1m152/153[0m [32m━━━━━━━━━━━━━━━━━━━[0m[37m━[0m [1m0s[0m 71ms/step - accuracy: 0.9460 - loss: 0.7293
Epoch 70: val_accuracy did not improve from 0.98757

Epoch 70: ReduceLROnPlateau reducing learning rate to 3.999999989900971e-07.
[1m153/153[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m14s[0m 86ms/step - accuracy: 0.9380 - loss: 0.7345 - val_accuracy: 0.9751 - val_loss: 0.6271 - learning_rate: 2.0000e-06
Epoch 70: early stopping
Restoring model weights from the end of the best epoch: 56.

12. Plot Akurasi dan Loss (Gabungan Semua Fase)

In [24]:

# Gabungkan history
acc = history_1.history['accuracy'] + history_2.history['accuracy'] + history_3.history['accuracy']
val_acc = history_1.history['val_accuracy'] + history_2.history['val_accuracy'] + history_3.history['val_accuracy']
loss = history_1.history['loss'] + history_2.history['loss'] + history_3.history['loss']
val_loss = history_1.history['val_loss'] + history_2.history['val_loss'] + history_3.history['val_loss']

# Batas antar fase
boundary_1 = len(history_1.history['accuracy']) - 1
boundary_2 = boundary_1 + len(history_2.history['accuracy'])

plt.figure(figsize=(16, 6))

plt.subplot(1, 2, 1)
plt.plot(acc, label='Training Accuracy', linewidth=2)
plt.plot(val_acc, label='Validation Accuracy', linewidth=2)
plt.axvline(x=boundary_1, color='gray', linestyle='--', alpha=0.7, label='Fase 2 start')
plt.axvline(x=boundary_2, color='black', linestyle='--', alpha=0.7, label='Fase 3 start')
plt.legend(fontsize=12)
plt.title('Training & Validation Accuracy', fontsize=14)
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.grid(alpha=0.3)

plt.subplot(1, 2, 2)
plt.plot(loss, label='Training Loss', linewidth=2)
plt.plot(val_loss, label='Validation Loss', linewidth=2)
plt.axvline(x=boundary_1, color='gray', linestyle='--', alpha=0.7, label='Fase 2 start')
plt.axvline(x=boundary_2, color='black', linestyle='--', alpha=0.7, label='Fase 3 start')
plt.legend(fontsize=12)
plt.title('Training & Validation Loss', fontsize=14)
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.grid(alpha=0.3)

plt.tight_layout()
plt.show()

13. Test Time Augmentation (TTA)

Menggunakan model terbaik (EMA) dan menerapkan augmentasi geometrik saat inferensi untuk meningkatkan akurasi.

In [25]:

# Load model terbaik
print(f"Loading best model from {checkpoint_path}...")
best_model = tf.keras.models.load_model(checkpoint_path, compile=False)

# Kumpulkan gambar test asli
raw_test_ds = tf.keras.utils.image_dataset_from_directory(
    test_dir, shuffle=False, batch_size=BATCH_SIZE, image_size=IMG_SIZE)

test_images = []
test_labels_true_raw = []
for images, labels in raw_test_ds.unbatch():
    test_images.append(images.numpy())
    test_labels_true_raw.append(labels.numpy())

test_images = np.array(test_images)
test_labels_true = tf.one_hot(np.array(test_labels_true_raw), NUM_CLASSES).numpy()

TTA_STEPS = 5
tta_predictions = []

for i in range(TTA_STEPS):
    aug_images = data_augmentation(test_images, training=True)
    aug_images = preprocess_input(aug_images)
    preds = best_model.predict(aug_images, batch_size=BATCH_SIZE, verbose=0)
    tta_predictions.append(preds)
    print(f"  TTA step {i+1}/{TTA_STEPS}")

mean_tta = np.mean(tta_predictions, axis=0)

test_preds = np.argmax(mean_tta, axis=1)
test_true = np.argmax(test_labels_true, axis=1)
tta_acc = np.mean(test_preds == test_true)

print(f"\nTest Accuracy (TTA {TTA_STEPS}x): {tta_acc*100:.2f}%")

Loading best model from /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras...
Found 1052 files belonging to 4 classes.

2026-06-11 17:27:25.283775: 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_2352', 36 bytes spill stores, 36 bytes spill loads

2026-06-11 17:27:26.774709: 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.
2026-06-11 17:27:26.931108: 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.
2026-06-11 17:27:27.736506: 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.
2026-06-11 17:27:27.900098: 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.
2026-06-11 17:27:28.454775: 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.
2026-06-11 17:27:28.605712: 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.
2026-06-11 17:27:29.033373: 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.
2026-06-11 17:27:29.200411: 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.
2026-06-11 17:27:30.126019: 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.
2026-06-11 17:27:30.281824: 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.
2026-06-11 17:27:30.478454: 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.
2026-06-11 17:27:30.636887: 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.

  TTA step 1/5
  TTA step 2/5
  TTA step 3/5
  TTA step 4/5
  TTA step 5/5

Test Accuracy (TTA 5x): 98.67%

14. Evaluasi — Classification Report & Confusion Matrix

In [26]:

print("\nClassification Report:\n")
print(classification_report(test_true, test_preds, target_names=class_names))

cm = confusion_matrix(test_true, test_preds)
plt.figure(figsize=(10, 8))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
            xticklabels=class_names, yticklabels=class_names)
plt.title('Confusion Matrix', fontsize=14)
plt.xlabel('Predicted', fontsize=12)
plt.ylabel('True', fontsize=12)
plt.tight_layout()
plt.show()

Classification Report:

              precision    recall  f1-score   support

 Bercak Daun       0.98      0.96      0.97       226
  Daun Sehat       1.00      1.00      1.00       325
  Hawar Daun       0.98      1.00      0.99       242
  Karat Daun       0.98      0.99      0.99       259

    accuracy                           0.99      1052
   macro avg       0.99      0.99      0.99      1052
weighted avg       0.99      0.99      0.99      1052

15. Simpan Model Akhir

In [27]:

# Simpan model training final (dengan EMA weights)
final_path = os.path.join(ckpt_dir, 'best_model.keras')
model.save(final_path)
print(f"Model saved to {final_path}")

# Simpan juga versi tanpa EMA untuk fallback
model.save(os.path.join(ckpt_dir, 'final_model.keras'))
print("Final model saved.")

Model saved to /home/asephs/ZeaVis-Edu/Machine_Learning/best_model/best_model.keras
Final model saved.

16. Export Model untuk Produksi

Setelah training selesai, jalankan pipeline ekspor secara berurutan:

1. SavedModel + TFLite

python save_model.py

Memuat best_model/best_model.keras, membangun arsitektur bersih (tanpa augmentasi), dan mengekspor ke:

model/saved_model/ — format produksi
model/model.tflite — untuk perangkat mobile/edge

2. ONNX (Rust ML Service)

python convert_onnx.py

Mengonversi SavedModel ke model/model.onnx untuk digunakan oleh Rust/Axum/ONNX Runtime.

3. TensorFlow.js (Web)

export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
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

Catatan: Pipeline ekspor terpisah dari notebook karena save_model.py membangun ulang arsitektur tanpa layer training (GaussianNoise, augmentasi) untuk produksi.

17. Model Card — ZeaVis Edu Corn Disease Classifier

Atribut	Detail
Nama Model	ZeaVis Edu — EfficientNetV2B0 Classifier
Versi	2.0
Arsitektur	EfficientNetV2B0 (Transfer Learning) + GaussianNoise + Conv2D(512) + Conv2D(256) + Dense(1024)
Framework	TensorFlow 2.x / Keras (float32)
Dataset	~6000-8000 gambar daun jagung (4 kelas)
Kelas	Bercak Daun, Hawar Daun, Karat Daun, Daun Sehat
Input	Gambar RGB 224×224 piksel
Output	Probabilitas per kelas (softmax)
Augmentasi	Flip, Rotation, Zoom, Translation, Contrast, Brightness, MixUp, CutMix, RandomErasing, GaussianNoise
Optimizer	AdamW + EMA + Label Smoothing 0.2
Training	3 fase: Head (lr=1e-3) → Partial FT (lr=1e-4) → Full FT (lr=5e-5)
Target Akurasi	≥95% test accuracy
Cara Pakai	Upload gambar daun jagung → model memprediksi kelas penyakit
Etika	Model ini hanya untuk tujuan edukasi/penelitian pertanian. Jangan gunakan sebagai satu-satunya alat diagnosis. Konsultasi dengan ahli pertanian tetap diperlukan.

18. Upload ke Hugging Face

Upload model, TFLite, ONNX, dan TF.js ke Hugging Face Hub repo zeavis-edu/corn-leaf-disease-classifier.

Persyaratan: Token Hugging Face harus disetel sebagai environment variable HF_TOKEN.

Di Colab: buka Secrets manager (ikon 🔑 di panel kiri), tambahkan secret dengan nama HF_TOKEN dan isi dengan token Anda, lalu aktifkan akses Notebook.
Di lokal: export HF_TOKEN=hf_... atau simpan di .env.

Jalankan cell berikut untuk memulai pipeline upload otomatis.

1.9 MiB Raw Blame History Unescape Escape