RF-DETR Num_classes mismatch warning

šŸ¤ Community Help
1

Hi! Iā€™m getting this warning when evaluating RF-DETR:

num_classes mismatch: pretrain weights has 0 classes, but your model has 1 classes reinitializing detection head with 0 classes

when I pass in the checkpoint_best_ema.pth from my training as the pretrained weights. But my model seems to be performing quite well. What might be the issue, ā€œpretrain weights has 0 classesā€ is very concerning?

2

Hi @mila88!
Unfortunately I am unable to support local evaluation of RF-DETR. Apologies for the inconvenience!!