search

HKUST-Aerial-Robotics / MVDepthNet

This repository provides PyTorch implementation for 3DV 2018 paper "MVDepthNet: real-time multiview depth estimation neural network"
GNU General Public License v3.0
308 stars 72 forks source link

About the depth of MVS test set #27

Closed TangRuijie closed 4 years ago

TangRuijie commented 4 years ago

Hi, Sorry for bothering again. In other issues and your paper you mentioned that the depth values of the DeMoN dataset are in metric scale. I checked the depth of the dataset and most of the data are normal. However, when I checked the test set of MVS dataset, I found something wrong with the depth values. To be specific, in the MVS test set, there are some samples of a chair: image The mean depth value of corresponding depth map is about 7.4. However, when it comes to samples of a building: image The mean depth value of corresponding depth map is only around 2.2. This is really strange and I wonder if you noticed this problem in your experiment? I am looking forward to your reply.

Best regards!

TangRuijie commented 4 years ago

When I compute the mean depth value I have already ignored the invalid pixels in the depth maps.

WANG-KX commented 4 years ago

Hi, Thanks for your interest. The source of the scale is from the relative pose. In other words, the scale is consistent with the relative pose. In MVS dataset, the relative poses are calculated via SfM methods (e.g. colmap) thus do not have correct scales. Thus, we cannot get the right scale. Kaixuan

TangRuijie commented 4 years ago

Thanks for you reply!