GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose

Zhichao Yin, Jianping Shi
3/6/2018
cs.CV

Official code: yzcjtr/GeoNet (735 stars, MIT, last commit 2019-02-19).

Abstract

We propose GeoNet, a jointly unsupervised learning framework for monocular depth, optical flow and ego-motion estimation from videos. The three components are coupled by the nature of 3D scene geometry, jointly learned by our framework in an end-to-end manner. Specifically, geometric relationships are extracted over the predictions of individual modules and then combined as an image reconstruction loss, reasoning about static and dynamic scene parts separately. Furthermore, we propose an adaptive geometric consistency loss to increase robustness towards outliers and non-Lambertian regions, which resolves occlusions and texture ambiguities effectively. Experimentation on the KITTI driving dataset reveals that our scheme achieves state-of-the-art results in all of the three tasks, performing better than previously unsupervised methods and comparably with supervised ones.

View on arXivView PDF

Code Implementations(1)

yzcjtr/GeoNetOfficial100%
MIT

Code for GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose (CVPR 2018)

735183PythonMar 12, 20187 years agoMIT

Similar Papers

Unsupervised Learning of Depth and Ego-Motion from Monocular Video Using 3D Geometric Constraints

91% match1 repo77.7k2018

DO3D: Self-supervised Learning of Decomposed Object-aware 3D Motion and Depth from Monocular Videos

89% match2024

RM-Depth: Unsupervised Learning of Recurrent Monocular Depth in Dynamic Scenes

88% match1 repo412023

Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video

88% match1 repo4882019

DCPI-Depth: Explicitly Infusing Dense Correspondence Prior to Unsupervised Monocular Depth Estimation

88% match1 repo142024

Discriminately Treating Motion Components Evolves Joint Depth and Ego-Motion Learning

88% match2025

Cite this paper

@article{yin2018geonet,
  title  = {GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose},
  author = {Zhichao Yin and Jianping Shi},
  year   = {2018},
  eprint = {1803.02276},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1803.02276}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1803.02276.svg)](https://codeofpaper.com/paper/1803.02276)

Discussion