It seems you are passing the path argument to args.out_path (line 332 in train.py). Is that the normal procedure?
Should that be commented out, in order to use the path argument in Multiscale dataset? Or is it expected that the dataset folders are inside the output_path?
Thanks!