fuxiao0719 / GeoWizard

[arXiv'24] GeoWizard: Unleashing the Diffusion Priors for 3D Geometry Estimation from a Single Image

Home Page:https://fuxiao0719.github.io/projects/geowizard/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

about video test?

zhanghongyong123456 opened this issue · comments

For video testing, flicker is very serious, what can be done to reduce flicker

Hi, the current model is for monocular estimation and due to the randomness of the initialization noise, it can not guarantee smoothness on the video sequence. However, we find that initializing the same seed for different frames and increasing the ensemble steps will enhance consistency to some extent. Besides, utilizing some post-processing techniques (optical flow) will
further improve depth smoothness. BTW, we are also working on improving video consistency during training, so please stay tuned.

Hi, the current model is for monocular estimation and due to the randomness of the initialization noise, it can not guarantee smoothness on the video sequence. However, we find that initializing the same seed for different frames and increasing the ensemble steps will enhance consistency to some extent. Besides, utilizing some post-processing techniques (optical flow) will further improve depth smoothness. BTW, we are also working on improving video consistency during training, so please stay tuned.

Thanks, Looking to solve video consistency