How chrispathway/covid-ct-classifier works
This repository is a notebook-based COVID-19 CT-slice classifier. Its main workflow prepares and splits labeled images, trains and evaluates a fine-tuned MobileNetV3-Large, saves a checkpoint, and uses trained weights for single-image predictions. The README documents these stages and capabilities; the only sampled source file is requirements.txt, so implementation-level call relationships cannot be verified.
Main entry points: classifier.ipynb Model training · mobilenet_v3_large.pth Trained weights
Components 12
Image data
- CT slices
- Dataset split
classifier.ipynb - Image transforms
classifier.ipynb - Training augmentation
classifier.ipynb
Model and training
- MobileNetV3 model
classifier.ipynb - Model training
classifier.ipynb - Trained weights
outputs/checkpoints/mobilenet_v3_large.pth
Evaluation and inference
- Split evaluation
classifier.ipynb - Slice prediction
classifier.ipynb - Class probabilities
Other components
- Notebook user
- Kaggle dataset
Connections 12
- Kaggle dataset → CT slices: provides images
README.md - CT slices → Dataset split: split by class
README.md - Dataset split → Image transforms: feeds transforms
README.md - Image transforms → Training augmentation: adds training crops
README.md - Notebook user → Model training: runs notebook
README.md - Model training → MobileNetV3 model: fine-tunes
README.md - Model training → Trained weights: saves weights
README.md - Model training → Split evaluation: evaluates splits
README.md - Notebook user → Slice prediction: provides slice
README.md - Trained weights → Slice prediction: loads weights
README.md - Slice prediction → MobileNetV3 model: runs model
README.md - Slice prediction → Class probabilities: returns probabilities
README.md