amazon-science / omni-detr

PyTorch implementation of Omni-DETR for omni-supervised object detection: https://arxiv.org/abs/2203.16089
Other
64 stars 6 forks source link

Custom Dataset #4

Open SayBender opened 2 years ago

SayBender commented 2 years ago

Thanks for the amazing work.

I like to know your recommendations on how to train on custom dataset with lower number of possible classes (say 3), Then the number of queries would also need be changed. I have tired this with Deformable DETR and it does not let me change number of queries even after removing the classification head. I wonder if you fixed this issue.

Any recommendation on what and how to change the code for custom dataset implementation is very appreciated.