problem regarding depth_refinement_loss

[bpwl0121]

Hi Felix,

MonoRec/model/loss_functions/monorec_loss.py

Line 334 in b2b3dec

mono_sdl = mask_mean(mono_sparsedl.detach(), mono_sparsedl_mask)

MonoRec/model/loss_functions/monorec_loss.py

Line 339 in b2b3dec

stereo_sdl = mask_mean(mono_stereodl.detach(), mono_stereodl_mask)

you use the .detach() method in the sdl at depth_refinement_loss, which cut the backpropagation, different from the depth_refinement_loss in the paper. Or did I misunderstand?

[Brummi]

Hi bpwl0121,

thank you for looking into the code so thoroughly, I appreciate it a lot!
I double checked and the code I used for the paper corresponds to the version in the paper.

But you seem to be correct here, I will look into it.
I suspect I may have introduced a mistake when refactoring the code before publishing it.

Best,
Felix