Question about stddev in Eq. (7)

[Harry-Zhi]

Hi, Very great work and cannot wait to read the codes of this paper.

1. When you compute second moments in function "part_map_to_mu_L_inv", I am not clear why we don't inverse stddev but instead computing following variables z,b,c,z. Could you please explain this part a bit? stddev calculation

2.Also, in order to extract appearance features, in the codes part map and processed colour features are added instead of being concatenated as mentioned in the paper. Is this minor difference leads to better performance and or there is a typo?

Thank you very much for the help. Looking forward to the full release.

Shuaifeng