search

MyNiuuu / MOFA-Video

[ECCV 2024] MOFA-Video: Controllable Image Animation via Generative Motion Field Adaptions in Frozen Image-to-Video Diffusion Model.
https://myniuuu.github.io/MOFA_Video
Other
588 stars 31 forks source link

Evaluation Code #45

Closed haoweiz23 closed 1 month ago

haoweiz23 commented 2 months ago

Is there a plan to release the evaluation code? How can I quantitatively evaluate the quality of video generation?

MyNiuuu commented 1 month ago

As mentioned in the paper, we select 1000 test videos from WebVid-10M for quantitative evaluation.

First, we generate the video using the first frame and the sampled sparse motion flow (codes for sparse motion flow sampling are provided in the training codes). Then, we calculate the quantitative metrics between the generated video and the ground truth (GT) video.