吕文玉

1. You can modify `loss_vfl` weight. https://github.com/lyuwenyu/RT-DETR/blob/0b6972de10bc968045aba776ec1a60efea476165/rtdetrv2_pytorch/configs/rtdetrv2/include/rtdetrv2_r50vd.yml#L73 2. No, they are for two separate tasks. 3. Reference 1.

Yes, you just need to register new backbone using `@register()` see details https://github.com/lyuwenyu/RT-DETR/blob/main/rtdetrv2_pytorch/src/nn/backbone/hgnetv2.py#L272 --- Then replace old one with your registered module name in config https://github.com/lyuwenyu/RT-DETR/blob/main/rtdetrv2_pytorch/configs/rtdetrv2/include/rtdetrv2_r50vd.yml#L13

And this line should adapt to specific bakcbone. https://github.com/lyuwenyu/RT-DETR/blob/0b6972de10bc968045aba776ec1a60efea476165/rtdetrv2_pytorch/configs/rtdetrv2/include/rtdetrv2_r50vd.yml#L29

Yes, I think you are right. One possible solution is to add an extra adaptation module. You can reference this paper. - [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/pdf/2203.16527)

可以看下变量`C`里面是不是有`inf`； 可以检查下框框格式是不是`coco`类型

> Y has inconsistent type tensor(double) (Triggered internally at ../torch/csrc/jit/serialization/export.cpp:1484.) 改啥了没；用的是最新的代码吗；贴一下你这个文件里用到`where`的地方；还有`onnx`和`torch`的版本 https://github.com/lyuwenyu/RT-DETR/blob/main/rtdetrv2_pytorch/src/zoo/rtdetr/rtdetrv2_decoder.py

下面应该还有一个`where`

这是我的 刚试了一下没问题；onnx版本和你不一样

https://github.com/lyuwenyu/RT-DETR/issues/95

这个`cvperception `是我的私有仓库；如果需要的话；我可以改造下这个脚本的依赖问题