ruili3 / dynamic-multiframe-depth

[CVPR 2023] Multi-frame depth estimation in dynamic scenes. -- Li, Rui, et al. "Learning to Fuse Monocular and Multi-view Cues for Multi-frame Depth Estimation in Dynamic Scenes".

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About the parameter invd_depth_min_max

WangXuCh opened this issue · comments

Excuse me, is your invd_depth_min_max fixed at (0.33, 0.0025) during the experiment? Does it need to be continuously updated through learning like self supervision?

Yes, invd_depth_min_max is fixed during training and inference as we mainly focus on the driving scenarios where the depth range is approximately the same. It's interesting to make invd_depth_min_max a trainable parameter especially when the scene depth range varies for different samples :)

Yes, invd_depth_min_max is fixed during training and inference as we mainly focus on the driving scenarios where the depth range is approximately the same. It's interesting to make invd_depth_min_max a trainable parameter especially when the scene depth range varies for different samples :)

Thank you for your reply! In fact, making invd_depth_min_max a trainable parameter is a common practice in self supervised scenarios.