Loading video...

Video Failed to Load

Go Home

Monocular depth estimation is “impossible” because one image can’t measure depth geometrically. Our iDisc #CVPR2023 can group pixels w/o supervision and learn depth inductive bias on groups. We get LiDAR-like (but denser) depth from single images! More:

52,234 views • 3 years ago •via X (Twitter)

10 Comments

AntiFisherYu001's profile picture
AntiFisherYu0012 years ago

You are a murderer

Adrian's profile picture
Adrian3 years ago

@ValueAnalyst1 @a_meta4 check it out

MinotaurOnLucy's profile picture
MinotaurOnLucy3 years ago

As an outsider, I have the following questions: What is the current state-of-the-art performance for out of distribution cases? If it is not good, will we see a foundational model that achieves good out of distribution performance in the near term, or would it be more mid term?

julien michot's profile picture
julien michot3 years ago

Why "impossible"? Here you get the most plausible depths.

Nikolaos Sarafianos's profile picture
Nikolaos Sarafianos3 years ago

This is great work! Did you test it on humans at various distances from the camera by any chance?

Fisher Yu's profile picture
Fisher Yu3 years ago

We test the method on street scenes that have people. However, it is indeed interesting to see whether we can use the method to estimate depth in human-centric scenes.

ζ Pedram ζ's profile picture
ζ Pedram ζ3 years ago

Neat idea, github?

Fisher Yu's profile picture
Fisher Yu3 years ago

github link: The full code will be released before CVPR 2023.

test bot's profile picture
test bot3 years ago

@Scobleizer 69

𝘃𝗿𝗹𝗹𝗿𝘃's profile picture
𝘃𝗿𝗹𝗹𝗿𝘃3 years ago

@Scobleizer Great work!

Related Videos