Appreciation for the Hardwork

[WaOnEmperoR]

I really appreciate this work. But in order to run this program successfully, I had to modify the following:

1. in utils/metrics.py, add .cuda() after every torch.eye(num_classes) -> torch.eye(num_classes).cuda()
2. change opt = parser.parse_args() to opt = parser.parse_known_args() in every files
3. to get the configuration parameters, call opt[0] instead of opt, e.g. get_loaders(opt[0])
4. Convert the contents of OUT folder in cddchangedetectiondataset to 8-bit binary image (0 and 255 only), and put it in the new folder OUT_binary

[qaz670756]

Thank you for your contribution. Since the code was released two years ago, it’s likely that differences in library versions are causing these issues. I really appreciate your modifications, as they are helpful for other users.