Hi, i’m trying to create a dataloader for my dataset, all is ok, but when I try to pass an array/list “codes” to MaskBlock the dataloader is instantiated and ignores one of my classes…
When I don’t pass the parameter (codes=None) the dataloader normally loads the mask. Any suggestions about what the problem might be?
This is happening when using “background” for the values 0 (position 0 in the list) and the value class 1 (position 1) is ignored. For now, the workaround seems to be working:
-> Sum 1 in all classes and create a fake category at position 1 of the list/array
… of course for this, need generate all mask again
In my case:
codes = np.array(['_background_', 'nucleus', 'A', 'B', 'out_of_focus'], dtype='<U17')
codes = np.array(['_background_', 'ignore', 'nucleus', 'A', 'B', 'out_of_focus'], dtype='<U17')