Siyuan Bian

hello @dldaisy , we found a bug in calculating the rotation matrix for spine3 in inference, and the updated code is uploaded in ``fix_wrong_rotmat`` branch. The bug is probably the...

@dldaisy Yes, the bug is due to the negative determinant of matrix, and you can refer to [ref1](http://www.kwon3d.com/theory/jkinem/rotmat.html) or [ref2](https://www.ece.uvic.ca/~bctill/papers/mocap/Challis_1995.pdf) for further information. But neither of the references make a...

If I understand correctly, maybe you can convert quaternions to axis-angle or rotation matrix, and then use the forward function of SMPL.

@lulindeng I met with similar problems before. After I switch to use pytorch==1.6.0, the problem disappears.

You can refer to #52

@monacv @foermozhu I encounter the problem too. The bug occurs because in some modes, pyrender produces images with RGB format instead of RGBA format as intended. You can just manually...

Thanks for your reply @xuchen-ethz . Yes, the script keeps running, and I am wondering whether the error influences the correctness of preprocessing.

In [demo code](https://github.com/Jeff-sjtu/HybrIK/blob/main/scripts/demo_image.py), we have already replaced opendr with pytorch3d.

> Hi @alexrichardson21, > > Yes, it is definitely possible. We rely on SPIN pretrained backbone. So, you would need to train SPIN with proper data augmentation. I plan to...

As shown [here](https://github.com/Jeff-sjtu/HybrIK/blob/9b8681dcf3c902dd5dacc01520ba04982990e1e2/scripts/demo_video.py#L252), `vertices` is the predicted human mesh vertices. You can rotate it by multiplying it by a rotation matrix on the left.