search

VITA-Group / Diffusion4D

"Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models", Hanwen Liang*, Yuyang Yin*, Dejia Xu, Hanxue Liang, Zhangyang Wang, Konstantinos N. Plataniotis, Yao Zhao, Yunchao Wei
https://vita-group.github.io/Diffusion4D/
214 stars 3 forks source link

evaluation set #7

Open jbji opened 2 months ago

jbji commented 2 months ago

hi. thanks for this amazing work and data release. I wonder what 20 samples exactly did this work use for evaluation, which is important for fair comparison on evaluation metrics.

It seems that it's possibly the last 20 samples (can I confirm it?) of the provided id list, but it would be better if an additional evaluation id list can be provided.

hw-liang commented 1 month ago

Hi, thank you for your interests in our work. The released ids are lexicographically ordered and contain both train and evaluation splits. We leave out 20 animal- and human-centered assets for evaluation. We will release the ids later.