Discrepancy in number of classes in the configuration file

[faizan1041]

I was trying to train the model on the configuration file provided here Config/cascade_mask_rcnn_hrnetv2p_w32_20e.py. I'm failing to understand the number of classes, considering bordered, borderless and cell, it should be 3+1 classes instead of 81 which is here. Can someone provide any explanation on this?

[marcelodiaz558]

Hello, I have the same doubt, did you figured it out?

[faizan1041]

Yes, that file is a sample from somewhere else, the real number of classes should be (num_of_classes + 1)

[DB11051998]

https://github.com/open-mmlab/mmdetection/issues/6340

I guess this might clear your doubts!!