I think what might happen is that you might have an invalid bbox, for example with width/height less than one, or outside of the image. This image will get filtered by the dataset.
Can you check that?
Thank you sincerely.I have solve it.But I had another problem.
RuntimeError: cuda runtime error (59) : device-side assert triggered at /pytorch/aten/src/THC/THCTensorMathCompare.cuh:82
.I hope to get your help.
You maybe have an index in your dataset that is outside of the range defined in NUM_CLASSES?
Thank you so much for your answer. I have solved this problem according to your guidance. Thank you sincerely.
Great to hear that you managed to fix your issue!