DW icon indicating copy to clipboard operation
DW copied to clipboard

A Dual Weighting Label Assignment Scheme for Object Detection

Results 18 DW issues
Sort by recently updated
recently updated
newest added

RT(确实没看懂,具体是在做什么,动机是啥

this can sovle issue https://github.com/strongwolf/DW/issues/13#issue-1251466442 '*=' is an inplace operation

In FCOS, each FPN level is responsible for predicting objects of different sizes. But in DW, each object is predicted by all FPN levels, why? ![image](https://user-images.githubusercontent.com/26771037/179434087-df006215-23c4-4bae-88fe-5660696d8a3f.png)

hi, we are interested in your work, and you are welcome to add dw's work to our yolov6 for even greater gains![https://github.com/meituan/YOLOv6](url) We actually tried it on yolov6n using the...

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation:

You've mentioned the Top-k method for selecting candidate bags in your paper, but it seems that there is only soft center prior method in this repo. How can I change...

thanks for your work, I wonder how dose DW perform on Yolo(v5)?

Hello, I want to port DW to the target tracking algorithm SiamCAR, there is a problem with the shape parameter gt_labels in the loss function, I don't know how to...

How to apply DW to two-stage detectors, such as fast-rcnn? Is DW applied to the rcnn stage of fast-rcnn?

In the class CenterPrior, there are definitions as follows: ``` self.mean = nn.Parameter(torch.zeros(num_classes, 2), requires_grad=False) self.sigma = nn.Parameter(torch.ones(num_classes, 2)+0.11, requires_grad=False) ``` So in DW, these two parameters do not need...