Same ap value for every label at the end of epoch

[sulebaynes]

Dear mmrotate community and contributors,

I get these kind of results at the end of epochs where the values for ap are identical for most of the labels.

I am using a custom dataset with 11 labels already converted to DOTA format prior to training (I tried redet, retina and roi so far - I get very similar results).
Any idea about what could be causing the same ap calculation for every label? It looks as if there is a systematic error which causes the calculation to result in the same way.

Best regards