Sebastian Janampa Rojas comments

Results 25 comments of


                                            Sebastian Janampa Rojas

[Bug] installation of both mmdet nor mmpose freezes

same

Visualize with GLB

Hi, I am also working on a problem similar to yours. Could you share the code to get the plot with the meshes? Also, I think you forgot to add...

where is Uncertainty minimal query selection?

Hi, I think it is this part https://github.com/lyuwenyu/RT-DETR/blob/14971513a5ef805edaf434af1a11403d2f1a9036/rtdetr_pytorch/src/zoo/rtdetr/rtdetr_criterion.py#L111. The authors mentioned they used it for the classification loss

Learning rate and learning rate scheduler

@lyuwenyu Why did you not use a learning rate drop as in previous detr models?

Clarification about the paper and Code for Uncertainty Query Selection

> 1- The AIFI model sounds like just the normal encoder of the original DETR, but I understand the contribution of the paper is the addition CCFF. But how is...

Clarification about the paper and Code for Uncertainty Query Selection

> 1- RT-DETR uses a single scale for the encoder, and so does DETR, so there is no gain in the encoder computational cost. 2- RT-DETR has an additional module...

Clarification about the paper and Code for Uncertainty Query Selection

Yes, sorry, I misspelled. What I meant is that RT-DETR uses one-layer encoder.

Question about discrite sampling in RT-DETRv2

Also, why is there a deployment constraint when torch.grid_sample is used?

Question about discrite sampling in RT-DETRv2

Thank you so much for your well-explained answer. If, for deployment, you used discrete sampling, why is the model not trained using torch.grid_sample(..., mode='nearest')

Why content = enc_topk_memory.detach() that memory need to be detach() ？

You can find the information in this [paper](https://arxiv.org/pdf/2203.03605), section 3.5