How dfilipeloja/breakhis works
This repository trains and evaluates binary classifiers for benign versus malignant breast histopathology images. The main workflow selects BreakHis images into training data, trains an AlexNet or pretrained VGG19 model, then loads a saved VGG19 model to classify test images and produce metrics and a confusion-matrix image. A separate demo script evaluates images arranged under a demo directory. The excerpts show training and evaluation code, but not the test-data injection script; its behavior and data flow are therefore not mapped.
Main entry points: train.py Training workflow · predict.py Test prediction · generate-dataset.py Dataset selection · models Saved models
Components 14
Image data
- BreakHis dataset
- Dataset selection
generate-dataset.py - Training images
train - Test images
test - Demo images
demo
Model training
- Training workflow
train.py - Model architectures
models.py - Saved models
models
Prediction and evaluation
- Test prediction
predict.py - Demo evaluation
demo.py
Results
- Classification reports
results/classification_reports - Confusion matrices
results - Training graphs
graphs.py
Other components
- Dataset user
Connections 17
- Dataset user → BreakHis dataset: obtains
README.md - Dataset user → Dataset selection: runs
README.md - BreakHis dataset → Dataset selection: selects images
generate-dataset.py - Dataset selection → Training images: copies images
generate-dataset.py - Dataset user → Training workflow: runs
README.md - Training workflow → Training images: loads batches
train.py - Training workflow → Test images: loads validation batches
train.py - Training workflow → Model architectures: creates model
train.py - Training workflow → Saved models: saves AlexNet
train.py - Training workflow → Training graphs: plots training
train.py - Dataset user → Test prediction: runs
README.md - Test prediction → Saved models: loads VGG19
predict.py - Test prediction → Test images: classifies
predict.py - Test prediction → Classification reports: writes report
predict.py - Test prediction → Confusion matrices: writes matrix
predict.py - Demo evaluation → Saved models: loads VGG19
demo.py - Demo evaluation → Demo images: classifies
demo.py