search

Daniil-Osokin / lightweight-human-pose-estimation-3d-demo.pytorch

Real-time 3D multi-person pose estimation demo in PyTorch. OpenVINO backend can be used for fast inference on CPU.
Apache License 2.0
653 stars 137 forks source link

UnpicklingError #76

Closed Jiyoung12345 closed 3 years ago

Jiyoung12345 commented 3 years ago

First of all, thank you for giving us a good code.

I ran the code and an error occurred, so I leave a question.

To run the demo, I wrote the following code in the terminal.

python demo.py --model human-pose-estimation-3d.pth --video 0

However, it gave the following error:

Cannot load fast pose extraction, switched to legacy slow implementation.

No CUDA device found, inferring on CPU Traceback (most recent call last): File "demo.py", line 59, in net = InferenceEnginePyTorch(args.model, args.device, use_tensorrt=args.use_tensorrt) File "C:\Users\Jiyoung Yoon\PycharmProjects\3DbodyPose\modules\inference_engine_pytorch.py", line 22, in init

checkpoint = torch.load(checkpoint_path, map_location='cpu')

File "C:\Users\Jiyoung Yoon\PycharmProjects\3DbodyPose\venv\lib\site-packages\torch\serialization.py", line 608, in load return _legacy_load(opened_file, map_location, pickle_module, pickle_load_args) File "C:\Users\Jiyoung Yoon\PycharmProjects\3DbodyPose\venv\lib\site-packages\torch\serialization.py", line 777, in _legacy_load magic_number = pickle_module.load(f, pickle_load_args) _pickle.UnpicklingError: could not find MARK

My computer os is windows, and since I don't have a graphic card, I was hoping to run it with a CPU.

I do not know the meaning of "could not find MARK", and I do not know how to fix it, so I leave a question.

If you know how to fix this error, please let me know. Thanks again for providing a good code.

Daniil-Osokin commented 3 years ago

Hi! Possibly model was not fully downloaded, can you please re-download model? Also you can check this issue.

Jiyoung12345 commented 3 years ago

It helped. I downloaded the files using "download zip", not "git clone". Thank you.

Daniil-Osokin commented 3 years ago

Ok, closing.