Centre de documentation
scientifique

Parallel cut pursuit for minimization of the graph total variation / Hugo Raguet (2019)  

Public Titre : Parallel cut pursuit for minimization of the graph total variation Type de document : Article/Communication Auteurs : Hugo Raguet, Auteur ; Loïc Landrieu , Auteur Editeur : Ithaca [New York - Etats-Unis] : ArXiv - Université Cornell Année de publication : 2019 Conférence : ICML 2019, Workshop on Learning and Reasoning with Graph-Structured Representations in International Conference on Machine Learning 15/06/2019 15/06/2019 Long Beach Californie - Etats-Unis Open Access Proceedings Importance : 6 p. Format : 21 x 30 cm Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image [Termes IGN] algorithme Cut Pursuit [Termes IGN] optimisation (mathématiques) Résumé : (auteur) We present a parallel version of the cut-pursuit algorithm for minimizing functionals involving the graph total variation. We show that the decomposition of the iterate into constant connected components, which is at the center of this method, allows for the seamless parallelization of the otherwise costly graph-cut based refinement stage. We demonstrate experimentally the efficiency of our method in a wide variety of settings, from simple denoising on huge graphs to more complex inverse problems with nondifferentiable penalties. We argue that our approach combines the efficiency of graph-cuts based optimizers with the versatility and ease of parallelization of traditional proximal. Numéro de notice : C2019-051 Affiliation des auteurs : LASTIG MATIS+Ext (2012-2019) Autre URL associée : vers ArXiv Thématique : IMAGERIE/INFORMATIQUE Nature : Poster nature-HAL : Poster-avec-CL DOI : 10.48550/arXiv.1905.02316 Date de publication en ligne : 07/05/2019 En ligne : https://graphreason.github.io/papers/10.pdf Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=93350

Supervized segmentation with graph-structured deep metric learning / Loïc Landrieu (2019)  

Public Titre : Supervized segmentation with graph-structured deep metric learning Type de document : Article/Communication Auteurs : Loïc Landrieu , Auteur ; Mohamed Boussaha , Auteur Editeur : Ithaca [New York - Etats-Unis] : ArXiv - Université Cornell Année de publication : 2019 Projets : 1-Pas de projet / Conférence : ICML 2019, Workshop on Learning and Reasoning with Graph-Structured Representations in International Conference on Machine Learning 15/06/2019 15/06/2019 Long Beach Californie - Etats-Unis Open Access Proceedings Importance : 15 p. Langues : Français (fre) Descripteur : [Vedettes matières IGN] Intelligence artificielle [Termes IGN] apprentissage profond [Termes IGN] graphe [Termes IGN] segmentation [Termes IGN] semis de points Résumé : (auteur) We present a fully-supervized method for learning to segment data structured by an adjacency graph. We introduce the graph-structured contrastive loss, a loss function structured by a ground truth segmentation. It promotes learning vertex embeddings which are homogeneous within desired segments, and have high contrast at their interface. Thus, computing a piecewise-constant approximation of such embeddings produces a graph-partition close to the objective segmentation. This loss is fully backpropagable, which allows us to learn vertex embeddings with deep learning algorithms. We evaluate our methods on a 3D point cloud oversegmentation task, defining a new state-of-the-art by a large margin. These results are based on the published work of Landrieu and Boussaha 2019. Numéro de notice : C2019-050 Affiliation des auteurs : LASTIG MATIS (2012-2019) Autre URL associée : vers ArXiv Nature : Poster nature-HAL : Poster-avec-CL DOI : 10.48550/arXiv.1905.04014 Date de publication en ligne : 19/05/2019 En ligne : https://graphreason.github.io/papers/4.pdf Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=92819