Since the customers are all Chinese speaking, I did not add English version.
All the bounding boxes in the samples are labeled by hands using https://github.com/tzutalin/labelImg
The train readMe and test readMe are in the repository
These are the outputs of small triangles detection in Flask
Config:
Results:
We used and modified the yolov5 model from here: https://github.com/ultralytics/yolov5