search

google / dynibar

Implementation of DynIBaR Neural Dynamic Image-Based Rendering (CVPR 2023)
https://dynibar.github.io/
Apache License 2.0
851 stars 40 forks source link

video result #1

Open kwea123 opened 1 year ago

kwea123 commented 1 year ago

For the video results on the project page. Are they all generated by the method? I feel like some are mixture of ground-truth and novel views. If it is the case, from a research perspective it would be better to place some marker on the frames that are generated so people can figure out what works well and what doesn't.

Just a suggestion. Thanks for your great work.

dynibar commented 1 year ago

Hi, all the frames in the website are generated by the method. Some of frames in the video do correspond to the original camera path of input so they are almost same as input (if this is the part you are wondering). But it's good suggestion, I would add text to indicate which part is original path and which part is novel views, Thanks again @kwea123