/monocular_depth

Self-supervised monocular depth estimation

Primary LanguagePython

This is a repo discribing our depth estimation method

Most code based on monodepth2.

Our work is accepted by CVMP2020, Constant Velocity Constraints for Self-Supervised Monocular Depth Estimation

Comparision Results

(all models trained in a self-supervised manner and test on a same metrics and testset)

Methods abs rel sq rel RMSE RMSE log D1
Monodepth2 (ICCV2019) 0.115 0.882 4.701 0.190 0.879
FeatDepth (ECCV2020) 0.104 0.729 4.481 0.179 0.893
PackNet (CVPR2020) 0.111 0.785 4.601 0.189 0.878
Fisheyedepthnet (ICRA2020) 0.117 0.867 4.739 0.190 0.869
Ours 0.112 0.816 4.715 0.190 0.880

New comparison:

Methods abs rel sq rel RMSE rmse log D1 D2 D3
Hr-depth(AAAI 2021) 0.109 0.792 4.632 0.185 0.884 0.962 0.983
hrnet-18 0.109 0.799 4.628 0.186 0.887 0.963 0.982
hrnet-18 + vc 0.107 0.767 4.553 0.185 0.889 0.963 0.982
hrnet-32 0.104 0.838 4.638 0.183 0.895 0.963 0.982

Demo

whentesting on non-rigid objects

API

  1. Train models
sh start2train.sh
  1. Evaluating depth
sh disp_evaluation.sh
  1. Testing a sample
sh test_sample.sh