Pixel-Pair Occlusion Relationship Map (P2ORM): Formulation, Inference and Application - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

Pixel-Pair Occlusion Relationship Map (P2ORM): Formulation, Inference and Application

Résumé

We formalize concepts around geometric occlusion in 2D images (i.e., ignoring semantics), and propose a novel unified formulation of both occlusion boundaries and occlusion orientations via a pixel-pair occlusion relation. The former provides a way to generate large-scale accurate occlusion datasets while, based on the latter, we propose a novel method for task-independent pixel-level occlusion relationship estimation from single images. Experiments on a variety of datasets demonstrate that our method outperforms existing ones on this task. To further illustrate the value of our formulation, we also propose a new depth map refinement method that consistently improve the performance of stateof-the-art monocular depth estimation methods.
Fichier principal
Vignette du fichier
P2ORM.pdf (8.19 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03133241 , version 1 (05-02-2021)

Identifiants

Citer

Xuchong Qiu, Yang Xiao, Chaohui Wang, Renaud Marlet. Pixel-Pair Occlusion Relationship Map (P2ORM): Formulation, Inference and Application. European Conference on Computer Vision, Aug 2020, Online, France. pp.690-708, ⟨10.1007/978-3-030-58548-8_40⟩. ⟨hal-03133241⟩
22 Consultations
53 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More