GeoGraph: Graph-Based Multi-view Object Detection with Geometric Cues End-to-End - INRIA - Institut National de Recherche en Informatique et en Automatique Accéder directement au contenu
Chapitre D'ouvrage Année : 2020

GeoGraph: Graph-Based Multi-view Object Detection with Geometric Cues End-to-End

Ahmed Samy Nassar
  • Fonction : Auteur
Stefano D’aronco
  • Fonction : Auteur
Jan D Wegner
  • Fonction : Auteur

Résumé

In this paper we propose an end-to-end learnable approach that detects static urban objects from multiple views, re-identifies instances, and finally assigns a geographic position per object. Our method relies on a Graph Neural Network (GNN) to, detect all objects and output their geographic positions given images and approximate camera poses as input. Our GNN simultaneously models relative pose and image evidence, and is further able to deal with an arbitrary number of input views. Our method is robust to occlusion, with similar appearance of neighboring objects, and severe changes in viewpoints by jointly reasoning about visual image appearance and relative pose. Experimental evaluation on two challenging, large-scale datasets and comparison with state-of-the-art methods show significant and systematic improvements both in accuracy and efficiency, with 2-6% gain in detection and re-ID average precision as well as 8x reduction of training time.
Fichier principal
Vignette du fichier
eccv2020.pdf (9.91 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03213841 , version 1 (26-09-2021)

Identifiants

Citer

Ahmed Samy Nassar, Stefano D’aronco, Sébastien Lefèvre, Jan D Wegner. GeoGraph: Graph-Based Multi-view Object Detection with Geometric Cues End-to-End. Computer Vision – ECCV 2020, pp.488-504, 2020, ⟨10.1007/978-3-030-58571-6_29⟩. ⟨hal-03213841⟩
283 Consultations
168 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More