search

baegwangbin / MaGNet

[CVPR 2022 Oral] Multi-View Depth Estimation by Fusing Single-View Depth Probability with Multi-View Geometry
MIT License
210 stars 15 forks source link

The meaning of predicted depth #1

Closed stunback closed 2 years ago

stunback commented 2 years ago

Hello, I'm curious about the meaning of model prediction. Does it represent a relative depth or absolute depth in meters?

baegwangbin commented 2 years ago

Hi,

The depth maps are in meters (i.e. absolute depth).

The output of MaGNet has a shape of (B, H, W, 2), where B is the batch size and H (and W) is the input height (and width). The first output channel is the mean depth, and the second channel is the standard deviation. They correspond to the mu and sigma of a Gaussian distribution.