Feng Li comments

Results 119 comments of


                                            Feng Li

Object Detection and Inference Image

I do not have the code for the per-category evaluation, you can make some modifications to the evaluator. We do have the visualization code in [visualizer.py](https://github.com/IDEA-opensource/DN-DETR/blob/main/util/visualizer.py)

How to trainmy own dataset？(not COCO)

You can transform your dataset to COCO format or write a new dataset class for your dataset.

How to trainmy own dataset？(not COCO)

Hey, as our code and dataloader largely follows [DETR](https://github.com/facebookresearch/detr) and [Deformable DETR](https://github.com/fundamentalvision/Deformable-DETR). They have been widely adopted in many custom dataset. you can refer to their issues about how to...

Sizes of tensors must match except in dimension 1.

It seems you set the wrong dimension. Please check your code.

Sizes of tensors must match except in dimension 1.

This parameter is directly inherited from DAB-DETR and we do not verify it.

Amazing work!! I want to konw if there is possible to release model weights on BaiDu or Tsinghua Cloud?

Hey, thank you for your suggestion, we have released the models on BadiDu.

Table 2, model params clarification

Thanks for your careful reading. It is a typo. We will update this number to `50M` in our next version.

Table 2, model params clarification

Yes. Mask DINO and DINO are slightly different in detection head sharing.

Mask/box denoising hparams

Hey, we use the same noise hyper-params as in DINO. It is now opensource, you can refer to its code for more details. Yes, we add noise to init anchors...

masks of instance segmentation

Which codebase did you use?