[Bug]:

[ategen3rt]

Before Reporting

• I have pulled the latest code of main branch to run again and the bug still existed. 我已经拉取了主分支上最新的代码，重新运行之后，问题仍不能解决。

• I have read the README carefully and no error occured during the installation process. (Otherwise, we recommand that you can ask a question using the Question template) 我已经仔细阅读了README上的操作指引，并且在安装过程中没有错误发生。（否则，我们建议您使用Question模板向我们进行提问）

Search before reporting

• I have searched the DAMO-YOLO issues and found no similar bugs. 我已经在issue列表中搜索但是没有发现类似的bug报告。

OS

Ubuntu

Device

Nvidia T4

CUDA version

12.2

TensorRT version

8.6.1.6

Python version

3.10

PyTorch version

2.0.1+cu117

torchvision version

2.0.1+cu117

Describe the bug

tensorRT model for tensorRT 8 outputs incorrect bounding boxes. Technically, since it's incorrectly interpreting the input tensor, it could be wrong out the rest of the output too.

python tools/converter.py -f configs/damoyolo_tinynasL45_L.py -c model.pth --batch_size 1 --img_size 1024 --trt --end2end

This was discussed in #102

It is fixed in PR #113 which changes box_coding from 1 (BoxCenterSize) to 0 (BoxCorner).
See https://github.com/NVIDIA/TensorRT/tree/release/8.6/plugin/efficientNMSPlugin for more information on the parameters.

To Reproduce

Run
python tools/converter.py -f configs/damoyolo_tinynasL45_L.py -c best.pth --batch_size 1 --img_size 1024 --trt --end2end --trt_eval

The evaluation is 0%.
I also use demo command to predict some images. The output show that the bounding box seem randomly.

Hyper-parameters/Configs

No response

Logs

No response

Screenshots

No response

Additional

No response