2022 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct), Oct 1, 2022
Augmented Reality (AR) appears as one of the most promising technologies for the next decades wit... more Augmented Reality (AR) appears as one of the most promising technologies for the next decades with a lot of applications in Industry 4.0. Unfortunately, most of existing systems are limited to small environments and running on a single device. To cope with these limitations, we present in this paper an AR Cloud solution for realtime crowd mapping. We propose an AR cloud service architecture that is able to 1) continuously relocalize AR devices in relation to the reference coordinate system of a common global 3D map in order to estimate the initial pose of the AR device and to correct tracking drift as soon as possible; 2) locally map in 3D the real environment surrounding each AR device; 3) fuse these 3D local maps with the global 3D map in order to keep it up-to-date, mandatory feature when real environment changes over time. Finally, we show experiments of our AR cloud platform at a large scale.
HAL (Le Centre pour la Communication Scientifique Directe), Nov 25, 2015
Les yeux sont l'une des caractéristiques les plus saillantes du visage humain. La localisation de... more Les yeux sont l'une des caractéristiques les plus saillantes du visage humain. La localisation de la pupille de l'oeil permet d'obtenir des informations importantes qui peuvent être utilisées dans plusieurs applications de vision par ordinateur. Plusieurs eye-trackers commercialisés aujourd'hui peuvent estimer avec précision sa position. Cependant les spécifications matérielles très complexes et l'environnement de l'utilisateur contrôlé (résolution suffisante de l'image de l'oeil, bonnes conditions d'éclairage, légères variations de la pose de la tête) rendent ces solutions impossibles à utiliser dans un environnement arbitraire. Dans cet article, nous présentons une approche basée sur les arbres de régression aléatoires projetés sur l'espace de Hough pour estimer la position de la pupille. Nous démontrons à partir de plusieurs évaluations sur des bases de données publiques références que notre approche est très robuste aux conditions d'éclairage, à l'échelle, aux mouvement des yeux et aux fortes variations de la pose de la tête, et qu'elle apporte une amélioration significative par rapport aux méthodes existantes de l'état de l'art. Eyes are one of the most salient feature of the human face. The location of the eye-pupil allows to access to important informations and can be used in several computer vision applications. Several commercial eye-trackers can estimate with a good accuracy the eye-pupil location, but need a complex hardware specifications with a controlled user environment (sufficient eye image resolution, good illumination, small head pose variations) making these solutions impossible to use on an arbitrary environment. In this paper, we present an approach based on Hough random forest, we demonstrate by several evaluations on challenging public datasets that our approach is very robust to illumination, scale, eye movements and high head pose variations and yields a significant improvement regarding to a wide range of state of the art methods Mots clé : Champs aléatoires, arbres de décision, regression, classification, détection de la pupille, détection du regard
Camera relocalization is needed in several applications such as augmented reality or robot naviga... more Camera relocalization is needed in several applications such as augmented reality or robot navigation. However, it is still challenging to have a both real-time and accurate method. In this paper, we present our hybrid method combing machine learning approach and geometric approach for real-time camera relocalization from a single RGB image. We introduce our sparse feature regression forest to improve the machine learning part. In our regression forest, we propose a novel split function, that uses a whole feature vector instead of classical binary test function to improve the accuracy of 2D-3D point correspondences. Moreover, we use sparse feature extraction (SURF features) to reduce time processing. The results indicate that our method is the only real-time hybrid method (50ms per frame). We also achieve results as accurate as the best state-of-the-art methods (hybrid methods) and outperform machine learning based and sparse feature based methods.
The network-based visualization of large photo-realistic 3D landscape and city models has receive... more The network-based visualization of large photo-realistic 3D landscape and city models has received significant attention over the last five years in the computer graphics and network community. To make such 3D geoweb services appealing to a large audience, 3D city models must reach a sufficient level of realism and faithfulness. The first problematic was the transmission and visualization of such
2017 IEEE International Symposium on Mixed and Augmented Reality (ISMAR-Adjunct), 2017
Mixed and augmented reality (MAR) is on the brink of large-scale consumer level commercialization... more Mixed and augmented reality (MAR) is on the brink of large-scale consumer level commercialization. Standards will be required for MAR to succeed and proliferate as an information media and new contents platform. Standards will enable the development of MAR system components able to interoperate through defined interfaces and hence will enable the development of end-to-end solutions built on system components easily plugged in together to achieve contents sharing and interoperability. The workshop will be a place to present existing standards and demonstrate how they could ease the adoption of MAR in many domains, a place to present recent initiatives in order to coordinate efforts and share requirements coming from the industry. The discussion will lay a foundation to many issues of standardization for MAR: proper subareas for standards and abstract levels, physical and environment object representation, content file format, calibration process, tracking and recognition, augmentatio...
2018 International Conference on 3D Vision (3DV), 2018
Camera relocalization is needed in several applications such as augmented reality or robot naviga... more Camera relocalization is needed in several applications such as augmented reality or robot navigation. However, it is still challenging to have a both real-time and accurate method. In this paper, we present our hybrid method combing machine learning approach and geometric approach for real-time camera relocalization from a single RGB image. We introduce our sparse feature regression forest to improve the machine learning part. In our regression forest, we propose a novel split function, that uses a whole feature vector instead of classical binary test function to improve the accuracy of 2D-3D point correspondences. Moreover, we use sparse feature extraction (SURF features) to reduce time processing. The results indicate that our method is the only real-time hybrid method (50ms per frame). We also achieve results as accurate as the best state-of-the-art methods (hybrid methods) and outperform machine learning based and sparse feature based methods.
2006 Tenth IEEE International Symposium on Distributed Simulation and Real-Time Applications, 2006
The ever increasing speed of Internet connections has led to a point where it is actually possibl... more The ever increasing speed of Internet connections has led to a point where it is actually possible for every end user to seamlessly share data on Internet. Peer-to-peer (P2P) networks are typical of this evolution. The goal of our paper is to show that thanks to self-adaptive assignment techniques, server-less P2P networks can efficiently deal with very large environments such
When dealing with real estate business, it is quite difficult for estate agents to make customers... more When dealing with real estate business, it is quite difficult for estate agents to make customers understand the potential and the volumes of free spaces. Thus, we propose an application that aims to solve these issues based on a laying out scenario in which a seller and a customer collaborate. As the roles of both users are different, we propose an asymmetric collaboration where the two users do not use the same interaction setup and do not benefit from the same interaction capabilities.
Eyes is one of the most salient feature of the human face. Some key-points such as pupil and corn... more Eyes is one of the most salient feature of the human face. Some key-points such as pupil and corners allow to access to important information which can be used in several computer vision applications. Eyes appearance is very sensible to head pose variations, illumination condition and resolution which decrease the existing method. We propose a robust method based on Hough regression trees evaluated on challenging public datasets.
Ces travaux de these portent sur la visualisation de scenes urbaines tridimensionnellestres vaste... more Ces travaux de these portent sur la visualisation de scenes urbaines tridimensionnellestres vastes et tres complexes a travers un reseau. L'objectif majeur est d'obtenir une epresentation adaptee a la transmission et visualisation d'un ensemble de bâtiments constituant un modele 3D urbain. Cette representation multi-resolution des bâtimentsd'une ville, appelee PBTree, se base sur une representation 2D1/2 modelisant les bâtimentsa l'aide de prismes (representation 2D1/2 issue des methodes fiables et robustes de modelisation non-assitee de villes). Cette representation multi-resolution a ete concue pour optimiser la transmission (compression, progressivite) et la visualisation (niveaux de details, adaptabilite au point de vue) de modeles urbains. La multi-resolution est tres bien adaptee au survol des villes. Cependant, une solution est proposee dans le cas d'une navigation au niveau du sol, et est basee sur un calcul de visibilite permettant de selectionner le niveau de resolution transmis et visualise
2022 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct), Oct 1, 2022
Augmented Reality (AR) appears as one of the most promising technologies for the next decades wit... more Augmented Reality (AR) appears as one of the most promising technologies for the next decades with a lot of applications in Industry 4.0. Unfortunately, most of existing systems are limited to small environments and running on a single device. To cope with these limitations, we present in this paper an AR Cloud solution for realtime crowd mapping. We propose an AR cloud service architecture that is able to 1) continuously relocalize AR devices in relation to the reference coordinate system of a common global 3D map in order to estimate the initial pose of the AR device and to correct tracking drift as soon as possible; 2) locally map in 3D the real environment surrounding each AR device; 3) fuse these 3D local maps with the global 3D map in order to keep it up-to-date, mandatory feature when real environment changes over time. Finally, we show experiments of our AR cloud platform at a large scale.
HAL (Le Centre pour la Communication Scientifique Directe), Nov 25, 2015
Les yeux sont l'une des caractéristiques les plus saillantes du visage humain. La localisation de... more Les yeux sont l'une des caractéristiques les plus saillantes du visage humain. La localisation de la pupille de l'oeil permet d'obtenir des informations importantes qui peuvent être utilisées dans plusieurs applications de vision par ordinateur. Plusieurs eye-trackers commercialisés aujourd'hui peuvent estimer avec précision sa position. Cependant les spécifications matérielles très complexes et l'environnement de l'utilisateur contrôlé (résolution suffisante de l'image de l'oeil, bonnes conditions d'éclairage, légères variations de la pose de la tête) rendent ces solutions impossibles à utiliser dans un environnement arbitraire. Dans cet article, nous présentons une approche basée sur les arbres de régression aléatoires projetés sur l'espace de Hough pour estimer la position de la pupille. Nous démontrons à partir de plusieurs évaluations sur des bases de données publiques références que notre approche est très robuste aux conditions d'éclairage, à l'échelle, aux mouvement des yeux et aux fortes variations de la pose de la tête, et qu'elle apporte une amélioration significative par rapport aux méthodes existantes de l'état de l'art. Eyes are one of the most salient feature of the human face. The location of the eye-pupil allows to access to important informations and can be used in several computer vision applications. Several commercial eye-trackers can estimate with a good accuracy the eye-pupil location, but need a complex hardware specifications with a controlled user environment (sufficient eye image resolution, good illumination, small head pose variations) making these solutions impossible to use on an arbitrary environment. In this paper, we present an approach based on Hough random forest, we demonstrate by several evaluations on challenging public datasets that our approach is very robust to illumination, scale, eye movements and high head pose variations and yields a significant improvement regarding to a wide range of state of the art methods Mots clé : Champs aléatoires, arbres de décision, regression, classification, détection de la pupille, détection du regard
Camera relocalization is needed in several applications such as augmented reality or robot naviga... more Camera relocalization is needed in several applications such as augmented reality or robot navigation. However, it is still challenging to have a both real-time and accurate method. In this paper, we present our hybrid method combing machine learning approach and geometric approach for real-time camera relocalization from a single RGB image. We introduce our sparse feature regression forest to improve the machine learning part. In our regression forest, we propose a novel split function, that uses a whole feature vector instead of classical binary test function to improve the accuracy of 2D-3D point correspondences. Moreover, we use sparse feature extraction (SURF features) to reduce time processing. The results indicate that our method is the only real-time hybrid method (50ms per frame). We also achieve results as accurate as the best state-of-the-art methods (hybrid methods) and outperform machine learning based and sparse feature based methods.
The network-based visualization of large photo-realistic 3D landscape and city models has receive... more The network-based visualization of large photo-realistic 3D landscape and city models has received significant attention over the last five years in the computer graphics and network community. To make such 3D geoweb services appealing to a large audience, 3D city models must reach a sufficient level of realism and faithfulness. The first problematic was the transmission and visualization of such
2017 IEEE International Symposium on Mixed and Augmented Reality (ISMAR-Adjunct), 2017
Mixed and augmented reality (MAR) is on the brink of large-scale consumer level commercialization... more Mixed and augmented reality (MAR) is on the brink of large-scale consumer level commercialization. Standards will be required for MAR to succeed and proliferate as an information media and new contents platform. Standards will enable the development of MAR system components able to interoperate through defined interfaces and hence will enable the development of end-to-end solutions built on system components easily plugged in together to achieve contents sharing and interoperability. The workshop will be a place to present existing standards and demonstrate how they could ease the adoption of MAR in many domains, a place to present recent initiatives in order to coordinate efforts and share requirements coming from the industry. The discussion will lay a foundation to many issues of standardization for MAR: proper subareas for standards and abstract levels, physical and environment object representation, content file format, calibration process, tracking and recognition, augmentatio...
2018 International Conference on 3D Vision (3DV), 2018
Camera relocalization is needed in several applications such as augmented reality or robot naviga... more Camera relocalization is needed in several applications such as augmented reality or robot navigation. However, it is still challenging to have a both real-time and accurate method. In this paper, we present our hybrid method combing machine learning approach and geometric approach for real-time camera relocalization from a single RGB image. We introduce our sparse feature regression forest to improve the machine learning part. In our regression forest, we propose a novel split function, that uses a whole feature vector instead of classical binary test function to improve the accuracy of 2D-3D point correspondences. Moreover, we use sparse feature extraction (SURF features) to reduce time processing. The results indicate that our method is the only real-time hybrid method (50ms per frame). We also achieve results as accurate as the best state-of-the-art methods (hybrid methods) and outperform machine learning based and sparse feature based methods.
2006 Tenth IEEE International Symposium on Distributed Simulation and Real-Time Applications, 2006
The ever increasing speed of Internet connections has led to a point where it is actually possibl... more The ever increasing speed of Internet connections has led to a point where it is actually possible for every end user to seamlessly share data on Internet. Peer-to-peer (P2P) networks are typical of this evolution. The goal of our paper is to show that thanks to self-adaptive assignment techniques, server-less P2P networks can efficiently deal with very large environments such
When dealing with real estate business, it is quite difficult for estate agents to make customers... more When dealing with real estate business, it is quite difficult for estate agents to make customers understand the potential and the volumes of free spaces. Thus, we propose an application that aims to solve these issues based on a laying out scenario in which a seller and a customer collaborate. As the roles of both users are different, we propose an asymmetric collaboration where the two users do not use the same interaction setup and do not benefit from the same interaction capabilities.
Eyes is one of the most salient feature of the human face. Some key-points such as pupil and corn... more Eyes is one of the most salient feature of the human face. Some key-points such as pupil and corners allow to access to important information which can be used in several computer vision applications. Eyes appearance is very sensible to head pose variations, illumination condition and resolution which decrease the existing method. We propose a robust method based on Hough regression trees evaluated on challenging public datasets.
Ces travaux de these portent sur la visualisation de scenes urbaines tridimensionnellestres vaste... more Ces travaux de these portent sur la visualisation de scenes urbaines tridimensionnellestres vastes et tres complexes a travers un reseau. L'objectif majeur est d'obtenir une epresentation adaptee a la transmission et visualisation d'un ensemble de bâtiments constituant un modele 3D urbain. Cette representation multi-resolution des bâtimentsd'une ville, appelee PBTree, se base sur une representation 2D1/2 modelisant les bâtimentsa l'aide de prismes (representation 2D1/2 issue des methodes fiables et robustes de modelisation non-assitee de villes). Cette representation multi-resolution a ete concue pour optimiser la transmission (compression, progressivite) et la visualisation (niveaux de details, adaptabilite au point de vue) de modeles urbains. La multi-resolution est tres bien adaptee au survol des villes. Cependant, une solution est proposee dans le cas d'une navigation au niveau du sol, et est basee sur un calcul de visibilite permettant de selectionner le niveau de resolution transmis et visualise
Uploads
Papers by Jérôme Royan