I've already setup the model and inputs as:

device(type='cuda', index=0)

model = attempt_load(args.weights, map_location=device) # load FP32 model
im = torch.zeros(1, 3, *imgsz).to(device) # image size(1,3,320,192) BCHW iDetection

But still met the following exeption:

Traceback (most recent call last):
File "D:\sdk\envs\smpl-py38-torch110-cu111\lib\site-packages\torch\onnx\utils.py", line 1529, in _export
graph, params_dict, torch_out = _model_to_graph(
File "D:\sdk\envs\smpl-py38-torch110-cu111\lib\site-packages\torch\onnx\utils.py", line 1172, in _model_to_graph
params_dict = _C._jit_pass_onnx_constant_fold(
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument index in method wrapper__index_select)
python-BaseException

Hello, this paper has been seen for a long time, it is really a good job！ when is it convenient for you to share the code? Sincerely look forward to it！

Thank you for sharing the code with us. I noticed that you have a journal version available at: https://arxiv.org/abs/2304.10765. To better follow your methodology, could you update the code specifically for the COCOHumanParts preparation? This would greatly help us understand your process more effectively. Thanks again.

Thank you very much for your excellent work. I came across the paper PBADET related to your team, which mentioned that the detection could be extended to animals. Could you please confirm if this project can achieve similar results?

hi, can you provide a script to convert your models to onnx format, thanks a lot !

Is the difference concerning face landmarks?
Do you have any plans to release pretrained weights of body/head or body/head/hand model?

i wanna know in which file is Association Decoding which is mentioned in your paper located?

Commands: python .\demos\video.py -p .\videos\test.mp4 --data .\data\JointBP_CrowdHuman_face.yaml --imgsz 1536 --weights .\weights\ch_face_m_1536_e150_best_mMR.pt --device 0 --start 0 --end -1 --thickness 2 --alpha 0.2 --save-size 540 --display

Hi, I have a question about your great work.

How did you determine the lambda value(Weight of the BPD Loss; 0.025 in hyp)?

Thank you in advance for the response!

Hi,
I was wondering if you plan to share your *.pt weights for quick testing.
By the way, thank you for your great work. It was the one I really waited for.