search

svip-lab / Indoor-SfMLearner

[ECCV'20] Patch-match and Plane-regularization for Unsupervised Indoor Depth Estimation
150 stars 24 forks source link

multi scale question #13

Closed zhangzaibin closed 2 years ago

zhangzaibin commented 2 years ago

Thank you for your excellent work. Why multi-scale training strategy is disabled in your code. Multi-scale loss works very well in Monodepth2.

niujinshuchong commented 2 years ago

Hi,

Good question! I don't remember the exact reason. Maybe simply because we extract superpixels in scale 0 and don't generate superpixels at higher scales.