BoMingZhao / GaussianPrediction

[SIGGRAPH Conference 2024] GaussianPrediction: Dynamic 3D Gaussian Prediction for Motion Extrapolation and Free View Synthesis
46 stars 1 forks source link

Questions about the paper #2

Closed bbangsik13 closed 3 months ago

bbangsik13 commented 3 months ago

Hi,

Thank you for the great work. I have some questions as follows:

1) In Table A, is motion extrapolation the same as motion prediction in Table 1? 2) On the project page, are the Hyper-NeRF Dataset Prediction Results all within the [0.8,1.0] range? The videos of scenes with different lengths are concatenated, which makes it confusing. 🥲

Thanks in advance for your help!

BoMingZhao commented 3 months ago

Hi, Thanks for your interest in our work.

  1. Yes, motion extrapolation and motion prediction are the same.
  2. In the paper, since the ground truth data is only available in the range [0-1], we train our model using the subset [0-0.8] and use the remaining [0.8-1] range for evaluation. However, on the project page, in order to display clearer prediction results, we extrapolate many more frames, although the training still uses the [0-0.8] range.

Hope these answer your questions.

bbangsik13 commented 3 months ago

I see, thanks a lot for your help! 👍