Questions about once-for-all property

[fushh]

Hi~ Thanks for your excellent work. I still have some questions about the once-for-all property that needs to be explained in the paper.

In terms of progressive refinement, I assume that the gains in performance come from the model ensemble rather than DDIM or diffusion training. We can regard each step in progressive refinement as an instantiation of a fixed-initial-box model, as the initial boxes in each step are totally random after DDIM and box renewal. This hypothesis can be validated by simply throwing away all boxes in box renewal and we gain 46.0AP with the released checkpoint diffdet_coco_res50_300boxes.pth and 5 refinement steps, which is the same as the one without modifying. The discussion in Issue 16 also shows useless of DDIM.

As for dynamic boxes, you'd better compare Deformable DETR + iterative refinement + two stage with yours since such kind of Deformable DETR does not use learnable queries. In my experiments, this Deformable DETR variant achieves 46.2AP, 46.9AP, 47.0AP, and 47.0AP with 100, 300, 500, and 1000 topk queries. Although the gains are minor, dynamic boxes do not degrade performance.

Please correct me if there is something wrong with these experiments. Hoping more insight analyses will be provided in the future. Many thanks!

[ShoufaChen]

Hi,

Thanks for your interest.

Your findings are interesting, and we are glad to explore more about them.

Would you mind providing more details about your experiments with Deformable DETR + iterative refinement + two stage?

[fushh]

You can run with the following command: python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --coco_path /YOUR_PATH/coco --resume ../r50_deformable_detr_plus_iterative_bbox_refinement_plus_plus_two_stage-checkpoint.pth --eval --with_box_refine --two_stage --num_queries 500 No other modification is needed.

[ShoufaChen]

Thanks for your information.