search

xy-guo / Learning-Monocular-Depth-by-Stereo

Learning Monocular Depth by Distilling Cross-domain Stereo Networks, ECCV18
https://arxiv.org/abs/1808.06586
MIT License
94 stars 8 forks source link

Does the camera baseline matter? #3

Closed kwea123 closed 5 years ago

kwea123 commented 5 years ago

In your experiments, you use only the scene flow dataset as synthetic data. I wonder if the camera baselines (and other intrinsic parameters) are the same across the three different scenarios (flying things, driving, monkaa)? In my case, I want to train the proxy stereo net with scene flow and other additional data, does that degrade the network's performance due to different camera settings?

xy-guo commented 5 years ago

For stereo matching, it does not matter if the baselines are different because the outputs are disparity values (not dependent on the intrinsic parameters).

kwea123 commented 5 years ago

Another question, have you tried training on more than 100 images (e.g. on the whole training set) in the supervised setting? If so, does it perform better?

xy-guo commented 5 years ago

Yes, I will perform better. But our paper mainly focuses on the cases when there is no ground truth or very little ground truth data.