Sublabel-Accurate Multilabeling Meets Product Label Spaces - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

Sublabel-Accurate Multilabeling Meets Product Label Spaces

Résumé

Functional lifting methods are a promising approach to determine optimal or near-optimal solutions to difficult nonconvex variational problems. Yet, they come with increased memory demands, limiting their practicability. To overcome this drawback, this paper presents a combination of two approaches designed to make liftings more scalable, namely product-space relaxations and sublabel-accurate discretizations. Our main contribution is a simple way to solve the resulting semi-infinite optimization problem with a sampling strategy. We show that despite its simplicity, our approach significantly outperforms baseline methods, in the sense that it finds solutions with lower energies given the same amount of memory. We demonstrate our empirical findings on the nonconvex optical flow and manifold-valued denoising problems.
Fichier principal
Vignette du fichier
paper.pdf (1.86 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03358467 , version 1 (29-09-2021)

Identifiants

  • HAL Id : hal-03358467 , version 1

Citer

Zhenzhang Ye, Bjoern Haefner, Yvain Quéau, Thomas Möllenhoff, Daniel Cremers. Sublabel-Accurate Multilabeling Meets Product Label Spaces. 43rd DAGM German Conference on Pattern Recognition, 2021, Bonn (online event), Germany. ⟨hal-03358467⟩
39 Consultations
96 Téléchargements

Partager

Gmail Facebook X LinkedIn More