News
Sunday September 30Master projects for the academic year 2012-2013
The PERCEPTION team is seeking four master students to work on the projects listed below. The projects can start any time between January 2013 and July 2013. The duration of each project is 4-6 months. Please contact Radu Horaud for more information.
List of Master projects:
Piece-wise decomposition of a depth map into planes for robot locomotion
Continuous Gesture Recognition from Stereo for Humanoid Robot
Human-robot interaction using faces, sound localization and voice recognition
Understanding gestures: robust audio-visual learning on the humanoid robot NAO
Thursday September 6An article about our work on social robots (in French)
Lors d’un cocktail, une personne dialogue avec un petit groupe. Soudain, c’est à vous qu’elle s’adresse. Vous engagez la conversation. Facile ? Des millions d’années d’évolution ont pourtant été nécessaires pour que votre cerveau adopte ce comportement social basique ! Ce problème d’attention sélective est appelé « cocktail party effect » par les psychologues. Le résoudre est un point décisif pour que des robots interagissent avec les humains.
Wednesday August 29Book on Time-of-flight cameras published by Springer
This book describes a variety of recent research into time-of-flight imaging. Time-of-flight cameras are used to estimate 3D scene-structure directly, in a way that complements traditional multiple-view reconstruction methods. The first two chapters of the book explain the underlying measurement principle, and examine the associated sources of error and ambiguity. Chapters three and four are concerned with the geometric calibration of time-of-flight cameras, particularly when used in combination with ordinary colour cameras. The final chapter shows how to use time-of-flight data in conjunction with traditional stereo matching techniques. The five chapters, together, describe a complete depth and colour 3D reconstruction pipeline. This book will be useful to new researchers in the field of depth imaging, as well as to those who are working on systems that combine colour and time-of-flight cameras.
Wednesday November 16Outstanding paper award at ICMI’11
Our article "Finding Audio-Visual Events in Informal Social Gatherings" received the "Outstanding Paper Award" (best paper) at the IEEE/ACM 13th International Conference on Multimodal Interfaces (ICMI’11), Alicante, Spain, November 2011. The paper is authored by Xavi Alameda-Pineda, Vasil Khalidov, Radu Horaud and Florence Forbes.
The paper addresses the problem of detecting and localizing audio-visual events (such as people) in a complex/cluttered scenario such as a cocktail party. The work is carried out within the collaborative European project HUMAVIPS .
Friday February 12New European project granted to PERCEPTION and MISTIS
The PERCEPTION (vision) and MISTIS (statistics) teams at INRIA Grenoble Rhone-Alpes, together with four other partners, have just been granted a new European project under the FP7-ICT Cognitive Systems and Robotics, the HUMAVIPS project: Humanoids with Auditory and Visual Abilities in Populated Spaces.
Abstract: Humanoids expected to collaborate with people should be able to interact with them in the most natural way. This involves significant perceptual, communication, and motor processes, operating in a coordinated fashion. Consider a social gathering scenario where a humanoid is expected to possess certain social skills. It should be able to explore a populated space, to localize people and to determine their status, to decide to join one or two persons, to synthetize appropriate behavior, and to engage in dialog with them. Humans appear to solve these tasks routinely by integrating the often complementary information provided by multi sensory data processing, from low-level 3D object positioning to high-level gesture recognition and dialog handling. Understanding the world from unrestricted sensorial data, recognizing people’s intentions and behaving like them are extremely challenging problems. The objective of HUMAVIPS is to endow humanoid robots with audiovisual (AV) abilities: exploration, recognition, and interaction, such that they exhibit adequate behavior when dealing with a group of people. Proposed research and technological developments will emphasize the role played by multimodal perception within principled models of human-robot interaction and of humanoid behavior. An adequate architecture will implement auditory and visual skills onto a fully programmable humanoid robot. An open-source software platform will be developed to foster dissemination and to ensure exploitation beyond the lifetime of the project.
Website: http://humavips.inrialpes.fr
Partners: INRIA Grenoble Rhone-Alpes (France, coordinator), The Czech Technical University in Prague (Czech Republic), Aldebaran Robotics (France), IDIAP Research Institute (Switzerland), and Bielefeld University (Germany).
Project duration: 1 February 2010 to 31 January 2013 (36 mohths)
Project budget: 3,432,000 euros
European funding: 2,636,000 euros
Thursday December 10Robotic perception, on purpose
A short article about the European POP project, coordinated by Radu Horaud, appeared in ICT results.
The article was also published by ScienceDaily.
See also a shorter article published by wired.co.uk
Friday December 12Workshop on computational and biological models for auditory, vision, and their interactions
Workshop organized in conjunction with the European project POP (Perception on Purpose)
Monday September 8CRS Industrial Prize awarded at BMVC’08
The CRS Industrial Prize, sponsored by Computer Recognition Systems, was awarded to:
Amaël Delaunoy, Emmanuel Prados, Pau Gargallo, Jean-Philippe Pons and Peter Sturm
for their paper entitled:
Minimizing the Multi-view Stereo Reprojection Error for Triangular Surface Meshes
Paper presented at the British Machine Vision Conference (BMVC’08), Leeds, UK, September 2008.
Wednesday July 30Best paper award at the AMDO’08 conference
The paper Inverse Kinematics using Sequential Monte Carlo Methods by Elise Arnaud and co-author Nicolas Courty received the best paper award and AMDO’08 (Conference on Articulated Motion and Deformable Objects).
Saturday July 12An interview with Richard Broadbridge, CEO of 4D View Solutions
4D View Solutions is a start-up company created in October 2007 and which commercializes real-time 3D video software based on the multiple-camera 3D reconstruction technologies developed by the PERCEPTION group.
Tuesday January 15The third VISIONTRAIN school: Understanding Behavior from Video Sequences, 9-14 March 2008, Les Houches
The third VISIONTRAIN winter school will be held at the Les Houches School of Physics, 9-14 March 2008. The school’ topic is: Understanding Behavior from Video Sequences.
The school is organized by professor Vasek Hlavac from the Czech Technical University, Prague.
For more information, please visit the following website:
Sunday November 254D View Solutions SAS
4D View Solutions (4DViews or 4DV) is a company recently co-founded by Richard Broadbridge with five members of the PERCEPTION group.
4DViews is specialized in the production of rich 3D visual content using a large number of cameras and videos. For more information on 4DViews’ technologies and products, please visit the company’s website:
Monday July 16PERCEPTION at SIGGRAPH’07 "Emerging Technologies"
Put any object into the interaction space, and it is instantaneously modeled in 3D and injected into a virtual world populated with solid and soft objects. Push them, catch them, and squeeze them...
Wednesday July 4PACV 2007 (workshop on Photometric Analysis For Computer Vision)
Emmanuel Prados and Peter Sturm (members of the PERCEPTION group) co-organize the first international workshop on Photometric Analysis For Computer Vision - PACV’07 . The workshop will take place in conjunction with ICCV’07 (Eleventh IEEE International Conference on Computer Vision) on Oct 14, 2007, Rio de Janeiro, Brazil.
Monday March 19Final evaluation of the OCETRE project
The ANR/RNTL OCETRE project addressed the problem of real-time 3-D reconstruction using multiple-cameras for augmented reality applications. It gathered three partners: Total Immersion SA, Thales Group SA, and INRIA Rhône-Alpes. The PERCEPTION and MOAIS group at INRIA developped a new 3-D reconstruction technology based on extracting the visual hull of an object from its silhouettes gathered with several cameras.
In the example below the three characters (from left to right: Nicolas, Clément, and Serge, members of the OCETRE consortium) were reconstructed in real-time using 5 cameras and 6 PCs. Then the geometric and texture information were passed to the D’fusion software which renders the characters in real-time within a virtual environment.
The result of D’fusion can be seen on this video: Clement, Pascal, and David.
Tuesday January 16The second VISIONTRAIN school: Computational and neurophysiological models for visual perception
The second VISIONTRAIN thematic school will take place at the Les Houches physics school, from 24 to 30 March 2007. The topic of this session will be dedicated to the study of both computational and neurophysiological models for visual perception. The following topics will be addressed:
Monday (March 24): Vision and representation,
Tuesday: Depth perception and stereopsis,
Wednesday: Object representation and categorization,
Thursday: Attention (linking sensing and perception),
Friday: Vision and learning.
Wednesday September 27Propositions de stage Mastère 2 Recherche 2006-2007: Analyse d’images et de vidéos
Les sujets de stage de Mastère 2 Recherche que nous proposons pour l’année universitaire 2006-2007 sont en ligne. Ces stages se déroulent à l’INRIA Rhône-Alpes (à Montbonnot près de Grenoble) dans l’équipe de recherche PERCEPTION.
Les sujets de stage portent sur l’analyse d’images et de vidéos pour la capture de mouvement, la segmentation de scènes complexes, le tracking temporel d’objets déformables, le calcul de formes tri-dimensionnelles à partir de la modélisation de l’éclairage, etc.
Une indemnité de stage est prévue.
La durée du stage dépend de l’établissement universitaire mais en général elle est de 3 à 4 mois.
Thursday September 21Elise ARNAUD joins the PERCEPTION project
Elise ARNAUD was just appointed assistant professor at UJF. She has joined the PERCEPTION project on September 1st 2006.
Friday June 16Visit of Jose Gaspar
Jose Gaspar, Associate Professor at the Instituto Superior Técnico at Lisbon, Portugal, is visiting the group for 3 weeks in June 2006.
Thursday May 18Final Program for 3DCINE workshop now online.
IEEE Workshop on Three-Dimensional Cinematography
8:30-9:15 Invited talk
Takeo Kanade and PJ. Narayanan: Historical Perspective of the 4D Virtualized Reality Project.
9:15-10:15 Papers
1. Ismail Oner Sebe, Suya You, Ulrich Neumann: Model-Driven Video-Based Rendering for Vehicles.
2. Zhigang Zhu and Hao Tang : Content-Based Dynamic 3D Mosaics.
10:15-10:45 Morning break
10:45-11:45 Papers
1. Georgios Litos, Xenophon Zabulis, Georgios Triantafyllidis: Synchronous Image Acquisition based on Network Synchronization.
2. Hansung Kim, Ryuuki Sakamoto, Itaru Kitahara, Kiyoshi Kogure: Cinematized Reality - Cinematographic 3D Video System for Daily Life Using Multiple Outer/Inner Cameras.
11:45-12:15 Invited paper
Adrian Hilton : Multiple camera studio production of actor performance.
12:15-1:30 Lunch
1:30-2:15 Invited talk
Abhijit Ogale and Yannis Aloimonos: The way of the future. Segmentation into surfaces: a compositional approach.
2:15-3:15 Papers
1. Takanori Senoh, Terumasa Aoki, Hiroshi Yasuda and Takuyo Kogure: Space-Sampling Method for 3d Cinemas.
2. Dejun Wang, Emanuel Prados, Stefano Soatto: Towards robust and physically plausible shaded stereoscopic segmentation for piecewise constant albedo.
3:15- 3:45 Afternoon break.
3:45-4:15 Invited Paper
Takeshi Naemura : Light Field Live With Thousands of Lenslets.
4:15-5:00 Invited talk
Masayuki Tanimoto: Free Viewpoint Television for 3D Scene Reproduction and Creation.
5:00 Conclusion of workshop
Wednesday April 12Programme of the workshop on the representation and use of prior knowledge in vision
The workshop on the representation and use of prior knowledge in vision (WRUKPV) will be held in Graz, Austria, on Saturday May, 13, and is organized in conjunction with ECCV 2006.
The scientific programme of the workshop has just been posted on the workshop’s website.
Wednesday March 8The 2005 activity report
The 2005 activity report is available on line here, or in pdf format here.
Tuesday March 7Winter School: Optimization methods in computer vision
The VISIONTRAIN Marie Curie network organizes it’s first thematic school at the Ecole de Physique des Houches, from 13 to 17 of March 2006.
Wednesday January 183DCINE’06 (Workshop on 3D Cinematography)
Remi Ronfard (member of the PERCEPTION group) co-organizes the first international workshop on 3-D Cinematography. The workshop will take place in conjunction with CVPR’06 (IEEE Conference on Computer Vision and Pattern Recognition) on June 22, 2006, New-York, NY.
Tuesday November 29The CYCLOPE Tracker at INTUITION2005
The CYCLOPE tracker was presented at the INTUITION2005, Paris, November 24,25. The INTUITION2005 international Workshop focussed on VR/VE & INDUSTRY: Challenges and opportunities. A POSTER describing the CYCLOPE system was also presented.
Monday November 28Int. Work. on the Representation and Use of Prior Knowledge in Vision
This workshop will be hold in conjunction with ECCV 2006, in Graz, Austria, on 13 May 2006 (Saturday). See details on the workshop’s website (scientific topics, call for papers, programme committee, paper submission information).
Friday October 14PhD position available
The PERCEPTION group seeks a strong PhD candidate to work on "The fusion of depth and monocular visual cues for building higher level scene descriptions"
Thursday September 29The MV Platform at ICCV 2005
This platform presents a scalable architecture to compute, visualize and interact in real time with 3D dynamic textured models of real scenes. This architecture is designed for mixed reality applications requiring such dynamic models, such as tele-immersion. The system is built upon of 3 main components: image acquisition, based on standard firewire cameras; model computation, based on a distribution scheme over a cluster of PC and using an optimal shape-from-silhouette algorithm ; model visualization. The distribution scheme ensures the scalability of the system, interactive frame rates and low latency.
Tuesday September 27Propositions de stage Mastère 2 Recherche : Analyse d’Images et de Vidéos
Les sujets de stage de Mastère 2 Recherche que nous proposons pour l’année 2005-06 sont en ligne. Ces stages se déroulent à l’INRIA Rhône-Alpes (à Montbonnot près de Grenoble).
Une indemnité de stage est prévue.
La durée du stage dépend de l’établissement universitaire mais en général elle est de 3 à 4 mois.
Friday September 23Workshop on Modeling People and Human Interaction (PHI’05)
The final program of the IEEE workshop on Modeling People and Human Interaction (PHI’05) is now available
The workshop will take place in Beijing, China, on Saturday, October 15, 2005, two days before the ICCV main conference, and will bring together researchers in vision, graphics, and cognitive sciences to assess the current state-of-the-art, identify known problems and generate ideas for future progress towards automatic modelling and interpretation of people’s movements, actions, and interactions using video.
This workshop is jointly organized by Remi Ronfard, Pascal Fua (EPFL,CH), Adrian Hilton (Univ. Surrey,UK), Daniel Gatica-Perez (IDIAP,CH) and Marian Stewart Bartlett (UCSD, USA).
Thursday September 22The CYCLOPE tracker at SIGGRAPH 2005
The CYCLOPE optical tracker was presented at SIGGRAPH 2005 (Los Angeles, July 31 - Aug. 4, INRIA stand N°1358). The demonstrator associates the CONTACT software (developed by Stéphane Redon from the I3D team) with the CYCLOPE tracker developed by Hervé Mathieu. CONTACT allows an optimal handling of collisions between virtual objects. CYCLOPE estimates the position and orientation of a real object at 30 frames per second. Therefore, a user can manipulate a real object that is tracked by CYCLOPE and which interacts with a virtual one.
Tuesday August 2List of VISIONTRAIN PhD selected candidates is available
The list of VISIONTRAIN PhD candidates that were selected for the 11 Marie-Curie fellowships is available since August 2, 2005. In case negociations fail with one candidate, the VISIONTRAIN committee established a reserve list as well.
Friday July 1PhD applications for VISIONTRAIN
The deadline for submitting an application for a PhD position with the VISIONTRAIN Marie Curie network approaches soon. For more details see the VISIONTRAIN’s website.
Friday June 24Emmanuel Prados joins the PERCEPTION project
Emmanuel Prados was just appointed "chargé de recherche" at INRIA Rhône-Alpes. He will join the PERCEPTION project in December 2005. Currently he is a post-doctoral fellow in the Computer Vision Laboratory at UCLA. He obtained his PhD degree in 2004 from Université de Nice Sophia-Antipolis. Emmanuel’s expertise is in the area of PDE and variational methods for computer vision. In particular, he developed a method for recovering depth from a single image using shape from shading.
Sunday May 22The 2004 activity report
The PERCEPTION project follows up the MOVI project. This is the latest activity report. Activity reports for the period 1994-2003 are archived here
