OGNI-DC: Robust Depth Completion with Optimization-Guided Neural Iterations

Yiming Zuo, Jia Deng

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

Depth completion is the task of generating a dense depth map given an image and a sparse depth map as inputs. In this paper, we present OGNI-DC, a novel framework for depth completion. The key to our method is “Optimization-Guided Neural Iterations” (OGNI). It consists of a recurrent unit that refines a depth gradient field and a differentiable depth integrator that integrates the depth gradients into a depth map. OGNI-DC exhibits strong generalization, outperforming baselines by a large margin on unseen datasets and across various sparsity levels. Moreover, OGNI-DC has high accuracy, achieving state-of-the-art performance on the NYUv2 and the KITTI benchmarks. Code is available at https://github.com/princeton-vl/OGNI-DC.

Original languageEnglish (US)
Title of host publicationComputer Vision – ECCV 2024 - 18th European Conference, Proceedings
EditorsAleš Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, Gül Varol
PublisherSpringer Science and Business Media Deutschland GmbH
Pages78-95
Number of pages18
ISBN (Print)9783031726453
DOIs
StatePublished - 2025
Event18th European Conference on Computer Vision, ECCV 2024 - Milan, Italy
Duration: Sep 29 2024Oct 4 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume15061 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference18th European Conference on Computer Vision, ECCV 2024
Country/TerritoryItaly
CityMilan
Period9/29/2410/4/24

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • General Computer Science

Keywords

  • Depth Completion
  • Optimization-inspired Design

Fingerprint

Dive into the research topics of 'OGNI-DC: Robust Depth Completion with Optimization-Guided Neural Iterations'. Together they form a unique fingerprint.

Cite this