Monocular videos

[aaryapatel007]

Hello,

Thanks for the amazing work. I was curious to know if GauSTAR will work on monocular videos. If so, how should one go about feeding the input to the model?

[chengwei-zheng]

Hi,

Currently, GauSTAR supports sparse-view input (e.g., 8 views), but it does not work with monocular video due to the assumption that most surfaces are visible to the cameras.
One potential approach to enable the current GauSTAR for monocular input is to use multi-view generation methods, such as CAT4D, to synthesize multi-view videos from monocular input.