Output of MoSh++ v.s. AMASS

[yufu-liu]

Hi, thanks for this amazing and useful work!

I just tried an interesting test. First, I generated SMPL parameters from the same MoCap data in AMASS, and then I compared generated .pkl with .npz in AMASS by observing and plotting the data. The result shows that their parameters are slightly different, but animatiton is similar. Can I understand this result is because the optimal parameters of MoSh++'s two objective functions are different? i.e. parameters in the released library are different to the parameters used for generating AMASS dataset