alevangel opened this issue 3 years ago · 0 comments
How many epochs did you train the deepcrack on your dataset for fine tuning and obtaining same results of the paper?