Repository Notes

This article will be used to record fracture detection. The overall process will be to first detect and then classify the vertebral body in the detection box to determine whether it is fractured or not.

• Original repository

TODO

see details

• Upload drr detection weights.
• add yolov5 detection predict.
• add detection training.

Quick Start

Install

pip install -r requirements.txt

Well classified according to the category.

Data

.
├── cut_drr
│   ├── AP
│   │   ├── fracture_images
│   │   └── normal_images
│   ├── LA
│   │   ├── fracture_images
│   │   └── normal_images
│   ├── all
│   │   ├── fracture_images
│   │   └── normal_images

Training

[Note] the pretrain model can download from the Original repository.

python train.py

Predict

python predict.py

Eval

python eval.py

Detection classify

you can choose run rtdetr_paddle or rtdetr_pytorch

python detection_classify_from_predict_bbox.py

grad cam vis

you can use to visualize classify model

python grad_cam_vis.py

Using examples