Display output media 1x736 with 2D keyfeature overlay

[Ranieri Santanchè]

Hi,

I am reprocessing data on a pc with two gpu's one (GeFORCERTX) 8Gb and a 16Gb cpu with a resolution of 1x736. Despite this with reprocessing it does not give me any OutputMedia_1x736. Why?

This is the code's answer:


Downloading calibration for Cam0
Downloading calibration for Cam1
Downloading calibration for Cam2

Processing Traslazione_caudale_02
The pose pickles for OpenPose 1x736 already exist in the database. We will download them to avoid re-running pose estimation
INFO:root:Load extrinsics for Cam0 - already existing
INFO:root:Load extrinsics for Cam1 - already existing
INFO:root:Load extrinsics for Cam2 - already existing
Running OpenPose for Cam0
Running OpenPose for Cam1
Running OpenPose for Cam2
Synchronizing Keypoints
Using handPunch sync function.
You tried to filter 60 Hz signal with cutoff freq of 500. Will filter at 30 instead.
You tried to filter 60 Hz signal with cutoff freq of 500. Will filter at 30 instead.
You tried to filter 60 Hz signal with cutoff freq of 500. Will filter at 30 instead.
Writing synchronized videos
INFO:root:Augmenting marker set
1/1 - 1s - 691ms/epoch - 691ms/step
1/1 - 1s - 832ms/epoch - 832ms/step
INFO:root:Running Inverse Kinematics
c72e7c90-f7b7-4488-8817-4838a3a4f6cc_syncd_Cam0.mp4
Result posted to S3.
c72e7c90-f7b7-4488-8817-4838a3a4f6cc_syncd_Cam1.mp4
Result posted to S3.
c72e7c90-f7b7-4488-8817-4838a3a4f6cc_syncd_Cam2.mp4
Result posted to S3.
Traslazione_caudale_02.json
Result posted to S3.
Result posted to S3.
Result posted to S3.
Result posted to S3.
Result posted to S3.
Result posted to S3.
Result posted to S3.
Downloading calibration for Cam0
Downloading calibration for Cam1
Downloading calibration for Cam2


Also, in your opinion is it possible to implement an increase in resolutionPose to 1x736_2scales?

Thank you and best regards,
Ranieri

[Matt Petrucci]

Hi Ranieri,

It looks like the code is skipping running the data through pose estimation becuase it is already in the 1x736 resolution:

"The pose pickles for OpenPose 1x736 already exist in the database. We will download them to avoid re-running pose estimation"

You should be able run at 1x736_2, see this comment here: https://github.com/stanfordnmbl/opencap ... ons.py#L93. We were able to get it to work with 8 GB.

Hope this helps,

[Ranieri Santanchè]

Hi Matt thanks for the answer,

I also tried 1x736_2 scales but it gives me the following error:

runcell('Process data.', 'C:/Users/Lab_B/Documents/MyRepos/opencap-core/Examples/reprocessSessions_Lab_Biomecc.py')
Processing 26b77263-0412-467f-98ec-8627917443f0

Processing neutral
INFO:root:Load extrinsics for Cam0 - already existing
INFO:root:Load extrinsics for Cam1 - already existing
INFO:root:Load extrinsics for Cam2 - already existing
It is unclear which settings were used for pose estimation. We will re-run pose estimation
Running OpenPose for Cam0
Running OpenPose for Cam1
Running OpenPose for Cam2
Synchronizing Keypoints
('Static trial failed', 'Video synchronization failed. Verify your setup and try again. \n A fail-safe synchronization method is for the participant to\n quickly raise one hand above their shoulders, then bring it back down. \n Visit https://www.opencap.ai/best-pratices to learn more about \n data collection and https://www.opencap.ai/troubleshooting for \n potential causes for a failed trial.', 'Traceback (most recent call last):\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\main.py", line 397, in main\n synchronizeVideos(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 876, in synchronizeVideos\n pointList, confList, nansInOutList,startEndFrameList = synchronizeVideoKeypoints(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 962, in synchronizeVideoKeypoints\n keypointList,confidenceList = zip(*[removeOccludedSide(keys,conf,footMkrs,confidenceThreshold,visualize=False) for keys,conf in zip(keypointList,confidenceList)])\nValueError: not enough values to unpack (expected 2, got 0)\n')
Downloading calibration for Cam0
Downloading calibration for Cam1
Downloading calibration for Cam2

Processing Gait
INFO:root:Load extrinsics for Cam0 - already existing
INFO:root:Load extrinsics for Cam1 - already existing
INFO:root:Load extrinsics for Cam2 - already existing
It is unclear which settings were used for pose estimation. We will re-run pose estimation
Running OpenPose for Cam0
Running OpenPose for Cam1
Running OpenPose for Cam2
Synchronizing Keypoints
('Dynamic trial failed.\nTraceback (most recent call last):\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\main.py", line 397, in main\n synchronizeVideos(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 876, in synchronizeVideos\n pointList, confList, nansInOutList,startEndFrameList = synchronizeVideoKeypoints(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 962, in synchronizeVideoKeypoints\n keypointList,confidenceList = zip(*[removeOccludedSide(keys,conf,footMkrs,confidenceThreshold,visualize=False) for keys,conf in zip(keypointList,confidenceList)])\nValueError: not enough values to unpack (expected 2, got 0)\n', 'Video synchronization failed. Verify your setup and try again. \n A fail-safe synchronization method is for the participant to\n quickly raise one hand above their shoulders, then bring it back down. \n Visit https://www.opencap.ai/best-pratices to learn more about \n data collection and https://www.opencap.ai/troubleshooting for \n potential causes for a failed trial.', 'Traceback (most recent call last):\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\main.py", line 397, in main\n synchronizeVideos(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 876, in synchronizeVideos\n pointList, confList, nansInOutList,startEndFrameList = synchronizeVideoKeypoints(\n File "C:\\Users\\Lab_B\\Documents\\MyRepos\\opencap-core\\utilsChecker.py", line 962, in synchronizeVideoKeypoints\n keypointList,confidenceList = zip(*[removeOccludedSide(keys,conf,footMkrs,confidenceThreshold,visualize=False) for keys,conf in zip(keypointList,confidenceList)])\nValueError: not enough values to unpack (expected 2, got 0)\n')

F0904 11:38:33.283108 9232 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
F0904 11:39:39.874032 10112 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
F0904 11:41:22.416467 2436 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
F0904 11:44:11.549532 5968 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
F0904 11:46:05.179584 15328 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***
F0904 11:47:42.149080 14808 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0) out of memory
*** Check failure stack trace: ***

Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.


With 1x736 configuration works but not with 1x736_2scales.
What could it be?


Thank you and best regards,
Ranieri

[Matt Petrucci]

Hi Ranieri,

It looks like the camera sync is failing. For the trials that you are trying to reprocess, did you try the hand punch in the air to synchronize?

[Ranieri Santanchè]

Hi Matt, thank you for the answer.

Yes, in this session I tried the hand punch in the air to synchronize the videos for the "Dynamic trials".
But, unfotunatelly it gaves me the same error.

I hope you can suggest other ways to reprocess to 1x736_2scales.

Best regards,
Ranieri

[Matt Petrucci]

Hi Ranieri,

Maybe to just take a step back, what issues were your running into with the data that you need to reprocess at the different resolution?

[Ranieri Santanchè]

Hi Matt,

for the data I have acquired in some cases, there is a flotation of the model, i.e. it is not exactly on the ground. And there is also excessive movement of the pelvis (tilt, list, and rotation).
With this data, I wanted to try your GRF estimation tool, but probably with these acquisitions, the simulations may not be completely reliable.

Thank you and best regards,
Ranieri

[Matt Petrucci]

Hi Ranieri,

Makes sense. What is the task that you are trying to capture?

By the way, you may want to consider applying for our virtual office hours program: https://restore.stanford.edu/2024/08/20 ... stions-10/

[Ranieri Santanchè]

Hi Matt,

I'm trying to capture a simple Gait overground. But when I simulate the GRF ( only with the right contact side) I obtain overestimated data ( peak of 1200 N counter 890 N recorded with the force plates).

I used hrnet for pose estimation and 120fps with 3 iPhones 12.

However, I will consult my team in order to apply to your program.

Bests regards,
Ranieri