Pytorch

OpenCV

ONNX runtime

CUDA >= 9.0

1.Train model with your datatset and save model weights (.pt file) using unet_train.py on supervisely.ly

2.Convert model weights to ONNX format using pytorch_to_onnx.py

3.Obtain real time inference using crack_det_new.py

Crack segmentation model files can be downloaded by clicking this link