Temporally-consistent version of depth estimation

yzxing87 / pytorch-deep-video-prior

[NeurIPS 2020] Blind Video Temporal Consistency via Deep Video Prior

118 stars 22 forks source link

Temporally-consistent version of depth estimation #7

Closed semel1 closed 2 years ago

semel1 commented 2 years ago

Is it possible to generate temporally consistent video from depth map sequence obtained by depth estimation (MIDAS) similar to "Blind Video Temporal Consistency project" https://github.com/lulu1315/BlindConsistency or https://github.com/nbonneel/blindconsistency?

ChenyangLEI commented 2 years ago

Yes, it is possible.

semel1 commented 2 years ago

Thanks for your kind reply. Is using "model colorization" with 50 epoch the proper way to go? (input - original color video, processed - depth estimation)

yzxing87 commented 2 years ago

You can try 'model colorization' or 'model dehazing'. Since IRT may not be necessary for your task.

semel1 commented 2 years ago

Great, thanks again. I'll try as soon as I get my GPU (currently I'm trying to test using CPU only, but it is way too slow)