Open johndpope opened 4 years ago
Hi, John,
Thanks for your inquiry. Our work is based on inertial measurement units, so we now have no cues to integrate it into graph-based works. But I think your question is quite interesting, if you have any idea of integrating the IMU-based work with graph-based pose estimation works that I could help you or some further questions of our work, please do not hesitate to contact me. I am very happy to communicate with you.
Best, Chunzhi
------------------ Original ------------------ From: "John D. Pope"<notifications@github.com>; Date: Thu, May 21, 2020 06:31 PM To: "ChunzhiYi/3-DOF-angle-estimation-based-IMUs"<3-DOF-angle-estimation-based-IMUs@noreply.github.com>; Cc: "Subscribed"<subscribed@noreply.github.com>; Subject: [ChunzhiYi/3-DOF-angle-estimation-based-IMUs] not an issue (#1)
were there any attempts to integrate this to openpose?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.
Sorry to dump all this
"OpenPose is limited to only detect the 2D position of certain keypoints" https://medium.com/@apofeniaco/p3d-body-detection-with-smplify-x-ced6c38871df
openpose -> fbx https://github.com/CMU-Perceptual-Computing-Lab/openpose/issues/672 The problem is - openpose can't do angle estimation - that's how I ended up here.
3D Keypoints Reconstruction from 2D Keypoints https://github.com/CMU-Perceptual-Computing-Lab/openpose/issues/1552
There is this -
Self-supervised Learning of Motion Capture https://arxiv.org/abs/1712.01337 https://github.com/htung0101/3d_smpl
you might loook into docker to try this out
I spent a week or two attempting to get all this to run on ubuntu - with this code base / docker - you can get it up and running in 30 minutes (of build time).
https://github.com/wells-wei-wei/smplify-x_in_docker
Hi, John,
Thanks for sharing this idea.
I think this idea would be practical for the "take-home" rehab, especially beneficial for quantifying patients' rehab progress at home. Extracting 3D skeletal information from monoscular vision, vectorizing them and then plugging into the angle calculation algorithm. Totally worth trying!
Best, Chunzhi
------------------ Original ------------------ From: "John D. Pope"<notifications@github.com>; Date: Sat, May 23, 2020 05:13 AM To: "ChunzhiYi/3-DOF-angle-estimation-based-IMUs"<3-DOF-angle-estimation-based-IMUs@noreply.github.com>; Cc: "Chunzhi Yi"<chunzhiyi.hit@foxmail.com>;"Comment"<comment@noreply.github.com>; Subject: Re: [ChunzhiYi/3-DOF-angle-estimation-based-IMUs] not an issue (#1)
Sorry to dump all this
"OpenPose is limited to only detect the 2D position of certain keypoints" https://medium.com/@apofeniaco/p3d-body-detection-with-smplify-x-ced6c38871df
openpose -> fbx CMU-Perceptual-Computing-Lab/openpose#672 The problem is - openpose can't do angle estimation - that's how I ended up here.
3D Keypoints Reconstruction from 2D Keypoints CMU-Perceptual-Computing-Lab/openpose#1552
There is this -
Self-supervised Learning of Motion Capture https://arxiv.org/abs/1712.01337 https://github.com/htung0101/3d_smpl
you might loook into docker to try this out I spent a week or two attempting to get all this to run on ubuntu - with this code base / docker - you can get it up and running in 30 minutes (of build time). https://github.com/wells-wei-wei/smplify-x_in_docker
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.
were there any attempts to integrate this to openpose?