Centre de documentation
scientifique

A model you can hear: Audio identification with playable prototypes / Romain Loiseau (2022)  

Public Titre : A model you can hear: Audio identification with playable prototypes Type de document : Article/Communication Auteurs : Romain Loiseau , Auteur ; Baptiste Bouvier, Auteur ; Yann Teytaut, Auteur ; Elliot Vincent, Auteur ; Mathieu Aubry, Auteur ; Loïc Landrieu , Auteur Editeur : Ithaca [New York - Etats-Unis] : ArXiv - Université Cornell Année de publication : 2022 Projets : READY3D / Landrieu, Loïc Conférence : ISMIR 2022, International Society for Music Information Retrieval Conference 04/12/2022 08/12/2022 Bengaluru Inde Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement du signal [Termes IGN] apprentissage automatique [Termes IGN] bruit (audition) [Termes IGN] onde acoustique [Termes IGN] prototype Résumé : (auteur) Machine learning techniques have proved useful for classifying and analyzing audio content. However, recent methods typically rely on abstract and high-dimensional representations that are difficult to interpret. Inspired by transformation-invariant approaches developed for image and 3D data, we propose an audio identification model based on learnable spectral prototypes. Equipped with dedicated transformation networks, these prototypes can be used to cluster and classify input audio samples from large collections of sounds. Our model can be trained with or without supervision and reaches state-of-the-art results for speaker and instrument identification, while remaining easily interpretable. The code is available at https://github.com/romainloiseau/a-model-you-can-hear Numéro de notice : P2022-006 Affiliation des auteurs : UGE-LASTIG+Ext (2020- ) Thématique : INFORMATIQUE Nature : Preprint nature-HAL : Préprint DOI : 10.48550/arXiv.2208.03311 En ligne : https://doi.org/10.48550/arXiv.2208.03311 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=101330

Online segmentation of LiDAR sequences: dataset and algorithm / Romain Loiseau (2022)  

Public Titre : Online segmentation of LiDAR sequences: dataset and algorithm Type de document : Article/Communication Auteurs : Romain Loiseau , Auteur ; Mathieu Aubry, Auteur ; Loïc Landrieu , Auteur Editeur : Berlin, Heidelberg, Vienne, New York, ... : Springer Année de publication : 2022 Projets : READY3D / Landrieu, Loïc Conférence : ECCV 2022, 17th European Conference on Computer Vision 23/10/2022 27/10/2022 Tel Aviv Israel Proceedings Springer Importance : pp 301 - 317 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie [Termes IGN] données lidar [Termes IGN] données localisées 3D [Termes IGN] jeu de données [Termes IGN] segmentation Résumé : (auteur) Roof-mounted spinning LiDAR sensors are widely used by autonomous vehicles. However, most semantic datasets and algorithms used for LiDAR sequence segmentation operate on 360∘ frames, causing an acquisition latency incompatible with real-time applications. To address this issue, we first introduce HelixNet, a 10 billion point dataset with fine-grained labels, timestamps, and sensor rotation information necessary to accurately assess the real-time readiness of segmentation algorithms. Second, we propose Helix4D, a compact and efficient spatio-temporal transformer architecture specifically designed for rotating LiDAR sequences. Helix4D operates on acquisition slices corresponding to a fraction of a full sensor rotation, significantly reducing the total latency. Helix4D reaches accuracy on par with the best segmentation algorithms on HelixNet and SemanticKITTI with a reduction of over 5× in terms of latency and 50× in model size. The code and data are available at: https://romainloiseau.fr/helixnet. Numéro de notice : C2022-043 Affiliation des auteurs : UGE-LASTIG+Ext (2020- ) Autre URL associée : vers ArXiv Thématique : IMAGERIE Nature : Communication nature-HAL : ComAvecCL&ActesPubliésIntl DOI : 10.1007/978-3-031-19839-7_18 Date de publication en ligne : 23/10/2022 En ligne : https://doi.org/10.1007/978-3-031-19839-7_18 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=101905

Representing shape collections with alignment-aware linear models / Romain Loiseau (2021)  

Public Titre : Representing shape collections with alignment-aware linear models Type de document : Article/Communication Auteurs : Romain Loiseau , Auteur ; Tom Monnier, Auteur ; Loïc Landrieu , Auteur ; Mathieu Aubry, Auteur Editeur : New York : Institute of Electrical and Electronics Engineers IEEE Année de publication : 2021 Autre Editeur : Ithaca [New York - Etats-Unis] : ArXiv - Université Cornell Projets : READY3D / Landrieu, Loïc Conférence : 3DV 2021, International Conference on 3D Vision 01/12/2021 03/12/2021 Londres online Royaume-Uni Proceedings IEEE Importance : pp 1044 - 1053 Format : 21 x 30 cm Note générale : bibliographie This work was supported in part by ANR project READY3D ANR-19-CE23-0007 and HPC resources from GENCI-IDRIS (Grant 2020-AD011012096). Langues : Anglais (eng) Descripteur : [Termes IGN] analyse de données [Termes IGN] apprentissage profond [Termes IGN] données localisées 3D [Termes IGN] modèle linéaire [Termes IGN] réseau neuronal profond [Termes IGN] segmentation [Termes IGN] semis de points [Termes IGN] transformation affine Résumé : (auteur) In this paper, we revisit the classical representation of 3D point clouds as linear shape models. Our key insight is to leverage deep learning to represent a collection of shapes as affine transformations of low-dimensional linear shape models. Each linear model is characterized by a shape prototype, a low-dimensional shape basis and two neural networks. The networks take as input a point cloud and predict the coordinates of a shape in the linear basis and the affine transformation which best approximate the input. Both linear models and neural networks are learned end-to-end using a single reconstruction loss. The main advantage of our approach is that, in contrast to many recent deep approaches which learn feature-based complex shape representations, our model is explicit and every operation occurs in 3D space. As a result, our linear shape models can be easily visualized and annotated, and failure cases can be visually understood. While our main goal is to introduce a compact and interpretable representation of shape collections, we show it leads to state of the art results for few-shot segmentation. Code and data are available at: https://romainloiseau.github.io/deep-linear-shapes Numéro de notice : C2021-036 Affiliation des auteurs : UGE-LASTIG+Ext (2020- ) Autre URL associée : vers ArXiv Thématique : INFORMATIQUE Nature : Communication nature-HAL : ComAvecCL&ActesPubliésIntl DOI : 10.1109/3DV53792.2021.00112 Date de publication en ligne : 03/12/2021 En ligne : https://doi.org/10.1109/3DV53792.2021.00112 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=98385