3d human pose estimation demos , pth converting problem

[hwang__seungjun]

Hi,

I encountered an error while trying to get "human-pose-estimation-3d-0001.xml".

I'm following this https://software.intel.com/en-us/forums/intel-distribution-of-openvino-toolkit/topic/844614

-----------------------------

========= Converting human-pose-estimation-3d-0001 to ONNX
Conversion to ONNX command: C:\Users\SJ\AppData\Local\Programs\Python\Python36\python.exe "C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\deployment_tools\open_model_zoo\tools\downloader\pytorch_to_onnx.py" "--model-path=C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\deployment_tools\open_model_zoo\tools\downloader\public\human-pose-estimation-3d-0001" --model-name=PoseEstimationWithMobileNet --model-param=is_convertible_by_mo=True --import-module=model "--weights=C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\deployment_tools\open_model_zoo\tools\downloader\public\human-pose-estimation-3d-0001/human-pose-estimation-3d-0001.pth" --input-shape=1,3,256,448 --input-names=data --output-names=features,heatmaps,pafs "--output-file=C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\deployment_tools\open_model_zoo\tools\downloader\public\human-pose-estimation-3d-0001/human-pose-estimation-3d-0001.onnx"

ERROR: Weights from
{}
 cannot be loaded for model {}! Check matching between model and weights
Error(s) in loading state_dict for PoseEstimationWithMobileNet:
        Unexpected key(s) in state_dict: "model.0.1.num_batches_tracked", "model.1.1.num_batches_tracked", "model.1.4.num_batches_tracked", "model.2.1.num_batches_tracked", "model.2.4.num_batches_tracked", "model.3.1.num_batches_tracked", "model.3.4.num_batches_tracked", "model.4.1.num_batches_tracked", "model.4.4.num_batches_tracked", "model.5.1.num_batches_tracked", "model.5.4.num_batches_tracked", "model.6.1.num_batches_tracked", "model.6.4.num_batches_tracked", "model.7.1.num_batches_tracked", "model.7.4.num_batches_tracked", "model.8.1.num_batches_tracked", "model.8.4.num_batches_tracked", "model.9.1.num_batches_tracked", "model.9.4.num_batches_tracked", "model.10.1.num_batches_tracked", "model.10.4.num_batches_tracked", "model.11.1.num_batches_tracked", "model.11.4.num_batches_tracked", "refinement_stages.0.trunk.0.trunk.0.1.num_batches_tracked", "refinement_stages.0.trunk.0.trunk.1.1.num_batches_tracked", "refinement_stages.0.trunk.1.trunk.0.1.num_batches_tracked", "refinement_stages.0.trunk.1.trunk.1.1.num_batches_tracked", "refinement_stages.0.trunk.2.trunk.0.1.num_batches_tracked", "refinement_stages.0.trunk.2.trunk.1.1.num_batches_tracked", "refinement_stages.0.trunk.3.trunk.0.1.num_batches_tracked", "refinement_stages.0.trunk.3.trunk.1.1.num_batches_tracked", "refinement_stages.0.trunk.4.trunk.0.1.num_batches_tracked", "refinement_stages.0.trunk.4.trunk.1.1.num_batches_tracked", "Pose3D.stem.0.bottleneck.0.1.num_batches_tracked", "Pose3D.stem.0.bottleneck.1.1.num_batches_tracked", "Pose3D.stem.0.bottleneck.2.1.num_batches_tracked", "Pose3D.stem.0.align.1.num_batches_tracked", "Pose3D.stem.1.bottleneck.0.1.num_batches_tracked", "Pose3D.stem.1.bottleneck.1.1.num_batches_tracked", "Pose3D.stem.1.bottleneck.2.1.num_batches_tracked", "Pose3D.stem.2.bottleneck.0.1.num_batches_tracked", "Pose3D.stem.2.bottleneck.1.1.num_batches_tracked", "Pose3D.stem.2.bottleneck.2.1.num_batches_tracked", "Pose3D.stem.3.bottleneck.0.1.num_batches_tracked", "Pose3D.stem.3.bottleneck.1.1.num_batches_tracked", "Pose3D.stem.3.bottleneck.2.1.num_batches_tracked", "Pose3D.stem.4.bottleneck.0.1.num_batches_tracked", "Pose3D.stem.4.bottleneck.1.1.num_batches_tracked", "Pose3D.stem.4.bottleneck.2.1.num_batches_tracked", "Pose3D.prediction.trunk.0.trunk.0.1.num_batches_tracked", "Pose3D.prediction.trunk.0.trunk.1.1.num_batches_tracked", "Pose3D.prediction.trunk.1.trunk.0.1.num_batches_tracked", "Pose3D.prediction.trunk.1.trunk.1.1.num_batches_tracked".

------------------------

How do I solve that?

[JAIVIN_J_Intel]

Hi,

I have just tried the same command on a windows machine and has successfully converted into onnx file. I used conda env to install Pytorch and to run this script. Also, make sure that the paths used to save the model has permission access.

Regards,

Jaivin