How arifsuz/cnn-vgg19-model-for-bovine-disease-classification works
Project type and purpose
• This is a data-science/ML pipeline project that implements a Convolutional Neural Network (VGG19) for cattle image classification. It is composed of Jupyter notebooks for dataset sorting, preprocessing, model training, and evaluation.
File structure
• Top-level files:
• LICENSE, README.md – documentation
• sortirDataset.ipynb – dataset organization
• preprocessing.ipynb – image conversion, resizing, augmentation
• model.ipynb – model definition, training, saving
• “hasil evaluasi model (tahap 1).txt” – evaluation summary
Extract key insights from the README
• Data flow: extract → organize into train/val/test → grayscale & RGB conversion → resize → augment
• Model: load pre‐trained VGG19, freeze layers, add pooling/dropout/dense layers, compile, train, save
• Evaluation: predictions → confusion matrix → classification report → accuracy/loss plots
Main components and interactions
A. Data Ingestion & Organization
• Raw Dataset (external ZIP)
• sortirDataset.ipynb organizes into folder hierarchy (train/val/test & class subfolders)
B. Preprocessing Pipeline
• preprocessing.ipynb
• OpenCV transforms: BGR→Gray, Gray→RGB, resize to 224×224
• Keras ImageDataGenerator for augmentation
C. Model Definition & Training
• model.ipynb
• Keras/TensorFlow VGG19 base (ImageNet weights)
• Custom head: GlobalAveragePooling2D, Dropout, Dense layers
• Compile (Adam, categorical_crossentropy), train on train/val generators
• Save model artifact (model_vgg19_cattle.h5)
D. Evaluation & Reporting
• model.ipynb & hasil evaluasi model.txt
• Predict on test set
• Compute confusion matrix (seaborn heatmap)
• Classification report (scikit‐learn)
• Plot accuracy/loss (Matplotlib)
Architectural patterns and design principles
• Linear pipeline architecture: each step feeds the next
• Modularity via separate notebooks for clear separation of concerns
• Reproducibility: artifacts (model file, evaluation metrics) stored for downstream use
1. Raw Dataset (ZIP)
2. Data Organizer (sortirDataset.ipynb)
3. Preprocessor (preprocessing.ipynb)
4. Model Trainer (model.ipynb)
5. Evaluator (model.ipynb + hasil evaluasi model.txt)
6. Artifacts Store (model_vgg19_cattle.h5 + evaluation plots/reports)
• → Preprocessor
• → Model Trainer
• → Evaluator
• → Artifacts Store
c. Technologies & libraries
• Data steps (blue rectangles)
• Model steps (green rectangles)
• Evaluation steps (orange rectangles)