FORTH-ModelBasedTracker / MocapNET

We present MocapNET, a real-time method that estimates the 3D human pose directly in the popular Bio Vision Hierarchy (BVH) format, given estimations of the 2D body joints originating from monocular color images. Our contributions include: (a) A novel and compact 2D pose NSRM representation. (b) A human body orientation classifier and an ensemble of orientation-tuned neural networks that regress the 3D human pose by also allowing for the decomposition of the body to an upper and lower kinematic hierarchy. This permits the recovery of the human pose even in the case of significant occlusions. (c) An efficient Inverse Kinematics solver that refines the neural-network-based solution providing 3D human pose estimations that are consistent with the limb sizes of a target person (if known). All the above yield a 33% accuracy improvement on the Human 3.6 Million (H3.6M) dataset compared to the baseline method (MocapNET) while maintaining real-time performance
https://www.youtube.com/watch?v=Jgz1MRq-I-k
Other
840 stars 135 forks source link

MocapNET2CSV to bvh worked, but got some errors #45

Closed xamxixixo closed 3 years ago

xamxixixo commented 3 years ago

Hello,

So I ran the MocapNET2CSV (from JSON files exported by Openpose). It worked and I had the out.bvh file. But the output animation is not completely right compared to the Openpose preview output. The body suddenly turned around 180 degrees and its position was lifted higher, just one time, then back to normal.

When I looked at the output log I saw some errors:

Failed @ loadCalibration (null) , color.calib

Loading Pose 0 -> dataset/poses/neutral.bvh  : Pose 0 -> dataset/poses/neutral.bvh  is corrupted
Failure
Loading Pose 1 -> dataset/poses/tpose.bvh  : Pose 1 -> dataset/poses/tpose.bvh  is corrupted
Failure
Loading Pose 2 -> dataset/poses/x.bvh  : Pose 2 -> dataset/poses/x.bvh  is corrupted
Failure
Loading Pose 3 -> dataset/poses/handsup.bvh  : Pose 3 -> dataset/poses/handsup.bvh  is corrupted
Failure
Loading Pose 4 -> dataset/poses/leftwave.bvh  : Pose 4 -> dataset/poses/leftwave.bvh  is corrupted
Failure
Loading Pose 5 -> dataset/poses/rightright.bvh  : Pose 5 -> dataset/poses/rightright.bvh  is corrupted
Failure
Loading Pose 6 -> dataset/poses/leftleft.bvh  : Pose 6 -> dataset/poses/leftleft.bvh  is corrupted
Failure
Loading Pose 7 -> dataset/poses/push.bvh  : Pose 7 -> dataset/poses/push.bvh  is corrupted
Failure
Loading Pose 8 -> dataset/poses/rightwave.bvh  : Pose 8 -> dataset/poses/rightwave.bvh  is corrupted
Failure
Failed to read recognized Poses
This is not fatal, but poses/gestures will be deactivated..

Are they errors that made the animation of out.bvh wrong? If so, how to fix?

Thank you.

AmmarkoV commented 3 years ago

Hello these are parts of the gesture recognition engine which are just an experimental feature.. The problem stems from the gestures recorded no longer corresponding to the main bvh file,

As stated in the output this is not a problem and will not effect 2D -> 3D functionality.

In any case its good you opened the issue so that I reduce needless console output spam

AmmarkoV commented 3 years ago

The messages won't be generated now except if a user explicitly calls the application using the --gestures which however is not currently working and will correctly give the error messages until the gesture bvh files are updated..