search

tiangexiang / OccNeRF

[ICCV 2023] Rendering Humans from Object-Occluded Monocular Videos
https://cs.stanford.edu/~xtiange/projects/occnerf/
MIT License
42 stars 1 forks source link

Question about evaluation #11

Closed arid29 closed 1 month ago

arid29 commented 3 months ago

Hi,

I noticed that when I run python eval.py --cfg ./configs/occnerf/zju_mocap/387/occnerf.yaml for obtaining the quantitative numbers, self.occlude is set to False. Can you please explain what's going on here?

tiangexiang commented 2 months ago

Hi, sorry for the late response. For simulated occlusions, we can have full evaluations of the entire human body by disabling the simulated occlusions. By doing this, we are able to comprehensively validate the ability of the models on both visible and occluded regions --- so we know if the model overfits on the visible regions or in fact learns the underlying human anatomy behind occlusions.

arid29 commented 1 month ago

thanks!