search

ardaduz / deep-video-mvs

Code for "DeepVideoMVS: Multi-View Stereo on Video with Recurrent Spatio-Temporal Fusion" (CVPR 2021)
MIT License
218 stars 29 forks source link

Cameras with Fixed Position #15

Closed Joywalker closed 2 years ago

Joywalker commented 2 years ago

Hello,

Really great work!

I've wanted to ask, would this model work based on the video input from 2 static cameras? Let's say I have two fixed in place cameras, and the only thing moving are the contents in front of the camera.

ardaduz commented 2 years ago

Hi,

This work is aimed at multi-view stereo with moving camera in static scenes, so it's not really suitable for your operating case out-of-the box. But the ideas presented in the paper like having a recurrent cell at the bottleneck to increase temporal consistency among two-camera stereo depth predictions hopefully inspires you when you're coming up with your solution.