Egocentric Video

Open rainbow979 opened this issue 2 years ago • 1 comments

Great work! I'm wondering whether the work can also predict the consistent depth of egocentric video.

Sep 17 '23 15:09 rainbow979

Hi! Thanks for your attention to our work. Our NVDS does not rely on pose estimation and is not restricted to fixed-camera videos. Any monocular videos including egocentric videos can be the input of NVDS. Some videos in our VDW training set are also egocentric, e.g., some first-person parkour videos from YouTube. Egocentric videos can be more challenging since they contain more intense camera motions. You can try some egocentric videos on our NVDS model and see if the results are satisfactory for you. I think the temporal depth consistency will also be significantly improved by NVDS.

Sep 18 '23 03:09 RaymondWang987