Question regarding derivative of dl_dmd in backward pass of diff_surfel_rasterization submodule

[harish-428]

Hello,

in line 361 of backward.cu, you calculate dl_dmd by using final_A and final_D, which are the final values of A and D. Shouldn't it be based on the values upto the current gaussian along the ray?

Thanks!

[hbb1]

Duplicate with this issue #28