Skip to content

Latest commit

 

History

History
61 lines (42 loc) · 1.89 KB

README.md

File metadata and controls

61 lines (42 loc) · 1.89 KB
Raw

SpindlesTracker

SpindleTracker is a automatically pipline for spindles detection, tracking and analysis.

1. Visualization demo

Tracking video

Skeletonize interface

2. Install

cd SpindleTracker
pip install -r requirements.txt
pip install -e .

3. Usage

Optimal: YOLOX-SP Training

  • Refer to our another repository: lizhogn/YOLOX-SP, you can train it on your own dataset.

1. Download the demo data and YOLOX-SP ONNX model

  • Demo file download: Google Drive (put the data to demo/data)

  • YOLOX-SP model download: Google Drive (put the weight to module/detection/weight)

2. Images Detection Demo

  • step2: run the following:
python demo_img.py --images path/to/input/image.png --model path/to/onnx/model

The results is saved in the image path (detection, mask and skeleton as shown in following).

3.2 TIF Input Demo

  • step1: import the ONNX model from YOLOX-SP and put it into module/detection/weight path (default path).

  • step2: run the following:

python demo_tif.py --images path/to/input/image.tif --model path/to/onnx/model --save_path path/to/save

The results is saved in save path. NOTE that the tif need to be double channel (red and green). The other input tif need to change the image propressing file and retrain the YOLOX-SP model.

3.3 GUI Interface Demo

The interface demo was built with gradio, first run the following:

python app/app.py

then, click the localhost link. Some images were listed in the example block, you can upload your own img.