questing about the 'deepcrack' result
smhotkdg opened this issue · 3 comments
smhotkdg commented
If I set the ground-truth pixel of the crack to be thicker than 1px, can the crack detected in the test get 1px or more?
qinnzou commented
If you use the pretrained model, the test results will be one pixel wide. If the crack ground truth are with width more than 1px, then you can trainyour own model with our codes. In such case, you can get results with larger crack width.
prinzROSE commented
Hello, there is no weight initialization step in your code, is there?
qinnzou commented
@prinzROSE training from scratch