Local Reprocessing using hrnet

New project for OpenCap, which is a new software package to estimate 3D human movement dynamics from smartphone videos. OpenCap strongly relies on OpenSim.
POST REPLY
User avatar
Jeevinthiran Karunagaran
Posts: 19
Joined: Sun Oct 27, 2019 6:10 am

Local Reprocessing using hrnet

Post by Jeevinthiran Karunagaran » Wed Dec 13, 2023 11:09 pm

Hello,

I am trying to locally reprocess some of my trials using hrnet to see if there are any differences in the results. But after instally pytorch, mmcv, mmengine and mmdet I still get an error.

Code: Select all

Traceback (most recent call last):
  File "c:\Users\jeevi\Desktop\opencap-core-main\main.py", line 296, in main
    videoExtension = runPoseDetector(
  File "c:\Users\jeevi\Desktop\opencap-core-main\utilsDetector.py", line 50, in runPoseDetector
    runMMposeVideo(
  File "c:\Users\jeevi\Desktop\opencap-core-main\utilsDetector.py", line 328, in runMMposeVideo
    from utilsMMpose import detection_inference, pose_inference
  File "c:\Users\jeevi\Desktop\opencap-core-main\utilsMMpose.py", line 14, in <module>
    from mmpose_inference import init_pose_model, init_test_pipeline, run_pose_inference, run_pose_tracking
  File "c:\Users\jeevi\Desktop\opencap-core-main\mmpose\mmpose_inference.py", line 6, in <module>
    from mmcv.runner import load_checkpoint
ModuleNotFoundError: No module named 'mmcv.runner'
How can I go about solving this?

User avatar
Scott Uhlrich
Posts: 124
Joined: Tue Jan 20, 2015 4:29 pm

Re: Local Reprocessing using hrnet

Post by Scott Uhlrich » Fri Dec 15, 2023 10:53 am

Hi - there currently is not a clean way to reprocess data locally with HRNet. You can try with the high resolution openpose settings.

User avatar
Jeevinthiran Karunagaran
Posts: 19
Joined: Sun Oct 27, 2019 6:10 am

Re: Local Reprocessing using hrnet

Post by Jeevinthiran Karunagaran » Thu Dec 21, 2023 10:08 am

Hello scott,

Thank you for the information. I will reprocess in openpose if needed.

Also is there a way for me to reprocess the data using a different model? I would like to reprocess my data using the shoulder model and not the original model. Should I use the changeSessionMetadata.py script to change the metadata to the LaiUhlrich2022_shoulder model and then use the reproecssSessions.py script to reprocess and get the results?

I tried doing this with one of my subjects and somehow the neutral pose disappeared and the dynamic trial failed :shock: . Could you help me check on this as I do not know how to recover the neutral trial. The session ID is :
03a7083d-cf62-4e17-b4a6-a11f94bd8b6b

Your help is much appreciated.

Regards,
Jeevin

User avatar
Jeevinthiran Karunagaran
Posts: 19
Joined: Sun Oct 27, 2019 6:10 am

Re: Local Reprocessing using hrnet

Post by Jeevinthiran Karunagaran » Fri Dec 22, 2023 1:05 am

I have found that following the steps of changing the metadata and reprocessing does indeed work at changing the model used. Its just that one trial that has a missing neutral pose that I need help with.

User avatar
Antoine Falisse
Posts: 422
Joined: Wed Jan 07, 2015 2:21 am

Re: Local Reprocessing using hrnet

Post by Antoine Falisse » Mon Jan 01, 2024 11:46 pm

Hey, I reprocessed your session. Take a look and let us know.

Best,
Antoine

POST REPLY