How robertsabaumance/osteval works
OstEval is an educational bone-X-ray estimation demo, but its planned upload-to-results application is not yet implemented. The repository currently provides three independent, stratified dataset-preparation scripts; each reads a disease dataset and writes train/validation/test CSVs. A separate pneumonia-classifier warm-up demonstrates transfer-learning practice and is explicitly unrelated to the final app. The README describes planned body-part routing, inference, Grad-CAM explanations, and image cleanup, but there is no source implementation to map. Edges reflect documented inputs and outputs; the planned backend is therefore omitted.
Main entry points: prepare_fracture.py Fracture split · prepare_oa.py OA grade split · prepare_op.py Osteoporosis split · pneumonia_classifier.py Pneumonia training
Components 11
Bone Data Preparation
- Fracture split
ml/data/prepare_fracture.py - Fracture splits
- OA grade split
ml/data/prepare_oa.py - OA splits
- Osteoporosis split
ml/data/prepare_op.py - Osteoporosis splits
Practice Training
- Pneumonia training
warmup/pneumonia_classifier.py
Other components
- FracAtlas data
- Knee OA data
- Knee osteoporosis data
- Pneumonia images
Connections 7
- FracAtlas data → Fracture split: reads labels
ml/data/prepare_fracture.py - Fracture split → Fracture splits: writes splits
ml/data/prepare_fracture.py - Knee OA data → OA grade split: reads images
ml/data/prepare_oa.py - OA grade split → OA splits: writes splits
ml/data/prepare_oa.py - Knee osteoporosis data → Osteoporosis split: reads images
ml/data/prepare_op.py - Osteoporosis split → Osteoporosis splits: writes splits
ml/data/prepare_op.py - Pneumonia images → Pneumonia training: loads folders
warmup/pneumonia_classifier.py