I am a distinguished professor in computer science at the Department of Mathematics and Informatics at Ecole Centrale de Lyon, France. I am also a senior member of the Institut Universitaire de France (IUF) through a chair of innovation starting from October 2022.
My general research interests are machine learning, computer vision and robotics, with a particular focus on the following themes:
Computer vision
face analysis and recognition in 2D and 3D;
image classification, object detection and semantic segmentation;
affective computing in images, audios, and videos;
Machine learning
deep learning,
transfer learning
unsupervised domain adaptation;
Robotics (find a presentation here)
Find my short bio here and my full CV here.
Find my publications at Google scholar, ResearchGate, dblp and hal.
Find my H-index and ranking according to Research.com Ranking here.
You have an excellent academic record! You are highly motivated for research in the field of machine learning, computer vision and robotics! Curiosity, open-mind, creativity, persistence, and collaborative-work ability are your personal skills, then apply for PhD or Postdoc positions with a short statement of research and detailed CV and join my research group to deal with cutting-edge research and application issues within the Liris lab at Ecole Centrale de Lyon, part of the elite of top 10 "Grande Écoles" in France offering access to excellent quality graduate and under-graduate students.
Opening of a PhD thesis proposal on Diffusion Model-Based Robot Visuomotor Policy Learning for General-Purpose and Multi-Task Robot Manipulation and Assembly Conditioned by Instructions;
Opening of an industrial PhD thesis proposal on occlusion handling in multi-object tracking in collaboration with Idemia, a world leading company on technologies of identification specialized in biometrics and video analytics;
A PhD position on rigid and deformable object modelling is opened for an expected starting date as early as possible;
A PhD position on a multi-task neural network for real time object detection and tracking on embedded systems for an expected starting date as early as possible;
A PhD position on multi-view reconstruction of deformable objects is opened for an expected starting date on 01-10-2022 within the ANR Rhino project;
A Postdoc position on disentangled latent manipulation learning for dexterous robot manipulation is opened for an expected starting date on 01-07-2022;
A Postdoc position on vision and touch empowered robot learning for dexterous bi-manual manipulation is opened for an expected starting date on 01-09-2021;
A PhD position is opened on molecular imaging analysis and machine learning within the framework of a cross-disciplinary collaboration between LBBE lab, Hospices Civils de Lyon, and the Imagine team, LIRIS lab at Ecole Centrale de Lyon.
A PhD position on Data Simulation, Machine Learning and Robotics is opened within the Chist-Era Learn Real project with a preferred starting date on September 2019; Closed !
A Postdoc position on data simulation, computer graphics and machine learning, extensible to 3 years, is opened within the FUI Pikaflex and the labcom Arès; Preferred starting date: January 5, 2019 !
A novel PhD position on Deep learning and multi-dimensional temporal signals is opened in collaboration with the Institut national des sciences & techniques nucléaires, CEA; closed !
A Postdoc position extensible to 2 years on machine learning and multimodal image indexing is opened within the ANR Alegoria project in collaboration with the Institut National de l'Information Géographique et Forestière (IGN); closed
NoPain: No-box Point Cloud Attack via Optimal Transport Singular Boundary will be presented at CVPR 2025, Nasheville, June 2025
Adversarial attacks exploit the vulnerability of deep models against adversarial samples. Existing point cloud attackers are tailored to specific models, iteratively optimizing perturbations based on gradients in either a white-box or black-box setting. Despite their promising attack performance, they often struggle to produce transferable adversarial samples due to overfitting the specific parameters of surrogate models. To overcome this issue, we shift our focus to the data distribution itself and introduce a novel approach named NoPain, which employs optimal transport (OT) to identify the inherent singular boundaries of the data manifold for cross-network point cloud attacks. Specifically, we first calculate the OT mapping from noise to the target feature space, then identify singular boundaries by locating non-differentiable positions. Finally, we sample along singular boundaries to generate adversarial point clouds. Once the singular boundaries are determined, NoPain can efficiently produce adversarial samples without the need of iterative updates or guidance from the surrogate classifiers. Extensive experiments demonstrate that the proposed end-to-end method outperforms baseline approaches in terms of both transferability and efficiency, while also maintaining notable advantages even against defense strategies.
The preprint can be found here: https://arxiv.org/abs/2503.00063
Code and model are available here: https://github.com/cognaclee/nopain
I’m so pleased to share that Prof. Huang Di, along with Prof. Yunhong Wang and Ass.Prof. Hongyu Yang, has been awarded the First Prize in Natural Science at the 2024 Wu Wenjun Artificial Intelligence Science and Technology Awards—the highest national recognition in intelligent science and technology in China.
Their research, "Efficient Representation Learning for Complex Visual Tasks", makes important theoretical and practical contributions by overcoming efficiency bottlenecks in visual representation learning across model design, data usage, and cross-domain transfer. The work has already seen impact across key sectors and has been praised for its simplicity, efficiency, and performance.
I am especially proud of this achievement because Prof. Huang Di completed his PhD under my supervision in 2010 within the Liris lab at Ecole Centrale de Lyon along with Prof.Yunhong Wang, and it’s been a joy to see him grow into a leading researcher. I also value the long-standing scientific collaboration I’ve had with the group of Prof. Yunhong Wang and Prof. Di Huang—their continued excellence and dedication to advancing AI is so fruitful in our joint research topics.
Congratulations to the entire team! This is a remarkable milestone and a strong step forward for the AI research community.
Pleased to share our last research entitled "Noise-Optimized Conditional Diffusion for Domain Adaptation (NOCDDA)", which has just been accepted in IJCAI, Montréal, August 2025
Domain adaptation is essential to enable models to generalize across shifting real-world conditions—without the burden of labeling every new domain. In this work, we tackle a core challenge in unsupervised domain adaptation (UDA) — the limited availability of high-confidence pseudo-labeled target samples. We introduce a conditional diffusion framework that bridges generative modeling and cross-domain decision-making, with a strong emphasis on class-aware noise optimization. The preprint can be found here: https://arxiv.org/abs/2505.07548v1
The International Joint Conference on Artificial Intelligence (IJCAI) is one of the premier global conferences in AI, bringing together top researchers from academia and industry since 1969. With a highly competitive review process and a strong focus on cutting-edge innovations across all areas of artificial intelligence, IJCAI has remained the premier conference bringing together the international AI community. This year’s edition will be held in Montréal, Canada, continuing its tradition of fostering high-impact AI research and collaboration.
Our former PhD student, Prof.Di Huang at Beihang University, nominated as Young scientific talent and China-France Outstanding Young Scientific Researcher 2024, is giving a seminar On 20/12/2024 at 14:00 to 15:00. Salle Szulkin, bâtiment E6, 1er étage, Ecully - Visio : https://ec-lyon-fr.zoom.us/j/97959739701
With the rapid development of the internet and sensing technology, different mo-dalities of data are emerging at an unprecedented speed. Multimodal data convey more comprehensive clues, but due to significant differences in their underlying mechanisms and forms of expression, and the complexity of the coupling relation-ships between modalities, how to utilize and mine the complementary information of multi-modal data has become an important research direction in the field. This talk focuses on visual perception based on multi-modal data and introduces the recent research work of the team of the presenter in this direction. The key discussion point is mainly on the fusion of texture and geometric information, involving data collected from various types of devices such as image sensors, depth sensors, LIDAR, for diverse tasks, including, facial analysis, robotic grasping, object detection, with the applications in industry, transportation, and other related areas.
With 100+ papers in the major academic journals and conferences such as IEEE-TPAMI, IJCV, CVPR, ICCV, ECCV, and 12,000+ citations and H-index 50, Prof.Di Huang has been among the top 2% world-wide AI scientist list since released in 2019 (by Stanford University and Elsevier).
« Interactive Data Visualization: Exploring and Struturing New Design Spaces», a Habilitation to Dirige Research of Ecole Centrale de Lyon, defended on 7-October-2024. Chair of the jury composed of Prof.Miriah MEYER (Linköping University-Sweden, Rapporteur), Dr.Emmanuel PIETRIGA (Research Director at INRIA, Rapporteur), Prof. Gennady ANDRIENIKO (Fraunhofer-IAIS Schloss Birlinghoven-Germany, Rapporteur), Dr. Michaël AUPETIT (Senior scientist at Qatar Computing Institut - Quatar, INRIA), Prof.Claudio SILVA (New York University - United States), and Prof.Angela BONIFATI (Université Claude Bernard Lyon 1).
A PhD thesis on “Taking advantage of new technologies, from deep learning to spatial omics: Towards a new clinically relevant morpho-molecular classification of pulmonary neuroendocrine tumors”, defended on 21-12-2023, in front of the jury composed of Prof.Delphine Maucort-Bouch (Chair, University Lyon 1 Claude Bernard), Prof.John Le Quesne (Rapporteur, University of Glasgow), Prof.Ruan Su (Rapporteur, University of Rouen), Dr.Garcia-Carbonero Rocio (Examinator, University of Madrid), Prof.Julien Mazières (Examinator, University of Toulouse 3), Dr.Lynnette Fernandez-Cuesta (Supervisor, IARC), Dr.Mathieu Foll (Co-supervisor, IARC), Prof.Liming Chen (Co-supervisor, ECL Liris)
Very honored to receive the 2022 award of applied research from FIEEC-Bpi France at the annual BIG (Bpi France Innovation Generation) event which gathered more than 65k participants and pretends to be the greatest business gathering in Europe. President Macron made the opening speech. This award rewards the research works that my group has carried out in partnership with Siléane, a leading French company in industrial robotics, to make their robots of manipulation more agile, more adaptive, more flexible. The award ceremony can be found on YouTube here.
Very pleased that our last collaborative research work on ImFace was presented today #CVPR2022. ImFace is a novel nonlinear 3D morphable face model with implicit neural representations mainly developed within Prof.Di Huang's group at #Beihang.
It builds two explicitly disentangled deformation fields to
model complex shapes associated with identities and expressions. It is a huge breakthrough compared to our last double linear 3D morphable model disentangled into identities and faciale expressions which appeared in #CVPR 2014. The paper can be found here ; the video presentation here ; and the code here.
Dimitri Gominski, our former doctoral student, currently in postdoc at the University of Copenhagen, won the Best paper award at the earthvision 2022 workshop in conjunction with the CVPR'2022 conference, with the paper entitled "Cross-dataset learning for generalizable land use scene classification". A paper which relates to his thesis work iunder co-supervision of Dr.Valérie Gouet-Brunet at the National Institute of Geography (IGN).
In this work, we lay the foundations towards an "universal" model for land-use mapping. We found that our data-driven approach leverages inter-dataset variety during training and intra-class similarities during testing yields good performance on 8 datasets, sometimes better than ad-hoc few-shot or cross-domain methods!
For more details, read our article here: https://openaccess.thecvf.com/content/CVPR2022W/EarthVision/html/Gominski_Cross-Dataset_Learning_for_Generalizable_Land_Use_Scene_Classification_CVPRW_2022_paper.html
A thesis funded by CSC and ANR on GAN-based Face Image Synthesis and its Application to Face Recognition. Started on 12 March 2018 and defended on 08-June-2022, in front of the jury composed of: Prof. Liming CHEN (Director of the thesis, ECL), Prof.Di HUANG (Co-Supervisor, Beihang University), Dr.Stéphane GENTRIC (IDEMIA, Examiner), Prof.Boulbaba BEN AMOR (Rapporteur, Telecom Lille), Dr.Sébastien Marcel(Rapporteur, Idiap), Prof.Alice Caplier (Chair, GIPSA Lab), Dr.Antitza Dantcheva (INRIA, Examiner).
Very pleased to learn yesterday and very honored to be nominated as a senior member at the IUF (Institut Universitaire de France) from October 2022 through the chair of innovation for the research project entitled "Look, Touch and Manipulate: Simulation empowered Self-supervised learning through self-play for dexterous robotic manipulation using Vision and Tactile Feeling", which has been judged as having high scientific value and strong potential for innovative development. #research #robotlearning #learning #IUF#
For further details, visit https://www.iufrance.fr/detail-de-lactualite/295.html
Historical visual sources are particularly useful for reconstructing the successive states of the territory in the past and for analysing its evolution. However, finding visual sources covering a given area within a large mass of archives can be very difficult if they are poorly documented. In the case of aerial photographs, most of the time, this task is carried out by solely relying on the visual content of the images. Convolutional Neural Networks are capable to capture the visual cues of the images and match them to each other given a sufficient amount of training data. However, over time and across seasons, the natural and man-made landscapes may evolve, making historical image-based retrieval a challenging task. We want to approach this cross-time aerial indexing and retrieval problem from a different novel point of view: by using geometrical and topological properties of geographic entities of the researched zone encoded as graph representations which are more robust to appearance changes than the pure image-based ones. Geographic entities in the vertical aerial images are thought of as nodes in a graph, linked to each other by edges representing their spatial relationships. To build such graphs, we propose to use instances from topographic vector databases and state-of-the-art spatial analysis methods. We demonstrate how these geospatial graphs can be successfully matched across time by means of the learned graph embedding. Find the all story in our last publication here: https://www.mdpi.com/2220-9964/11/2/97
A thesis funded by the Commissariat A l’Energie Atomique (CEA). PhD thesis on Deep Learning applied to multidimensional temporal signals. Started on 14-Dec-2018, and defended on 17-Dec-2021, in front of the jury composed of Prof.Liming Luke Chen (Chair, Ulster University), Dr.Hongying Meng (Rapporteur, reader at Brunel University London), Prof.Zhao Xi (Rapporteur, Xi’an Jiaotong University), Dr.Alice Othmani (Associate Prof, University of Paris Est), Dr. Marielle Malfante (Research engineer, CEA-LIST/DSCIN/LIIM), Prof.Liming Chen (Thesis director, ECL Liris), M.Andreas Vassilev (Co-supervisor, CEA/DSIS/SCSE/LSCM).
Deep neural networks have revolutionized Machine Learning, completely reshaping several domains of research in a mere decade. The most impressive changes were for domains like computer vision, for which deep learning outclassed the previous approaches based on handcrafted features. For instance, to the current day (end of 2021), the most baseline approach to extract features from a RGB image is to use a neural network trained on the popular image classification task ImageNet. More generally, in the most popular domains (Computer Vision, NLP, \textit{etc.}), there is a great deal of literature, good practices, and pretrained models accessible with a few lines of Python code. However, not all problems are as crowded as computer vision.
In some cases, a single laboratory can deal with multiple types of sensors recording temporal signals (accelerometers, strain sensors, GPS signals, physiological sensors) to perform Machine Learning on, in the span of a few months. For many unnoticed tasks, deep neural networks require considerable work: constitute a database, choose the type of preprocessing, hyperparameter selection, choice of an encoding, or even sensor choice, depending on the problem. We will take the place of a practitioner and review the most common choices we would have to make in order to make deep neural networks work with temporal signals. In each case, we will give indications the best choice to make, according to our experiments and/or the literature. Most of our experiments will focus on Transport Mode Detection, but we will use the literature to distinguish between the conclusions that only apply to our problem, and the affirmations that generalize elsewhere.
The defense of my 43rd thesis is under way...a thesis supported by ANR through the #Alegoria project and #DGA on Generalizable features and image search for multi-source interconnection and analysis by Dimitri Gominski, in front of the jury composed of Prof. Peter BELL (Rapporteur, Friedrich-Alexander Universität, Erlangen-Nürnberg, Germany), Prof. Philippe JOLY (Rapporteur, Université Paul Sabatier, Toulouse, France), Prof. Jantien STOTER (Chair, Delft University of Technology, Netherland), Prof. Dimitris SAMARAS (Stony Brook University, USA), Dr. Valérie GOUET-BRUNET (Supervisor, Senior researcher, Université Gustave Eiffel, France), Prof. Liming CHEN (Co-supervisor, École Centrale Lyon, France)
“Giving Robots Human Dexterity: simulation empowered object instance aware self-supervised learning for adaptive bin-picking”, keynote speech at The International Conference on Emerging Techniques in Computational Intelligence, http://www.ietcint.com/, technically co-sponsored by the IEEE Computational Intelligence Society, Ecole Centrale School of Engineering, Mahindra University, Hyderabad, India, on Aug 25-27, 2021
We are pleased that our Aristotle project has been just accepted within the 1st Franco-German joint call on AI. It teams up my group with Prof.Jan Peters' group at Technische Universität Darmstadt.
Dexterous manipulation has been a long-standing challenge in AI and robotics. Already Aristotle noted that the hand is the “tool of tools”, and Anaxagoras held that “man is the most intelligent of the animals because he has hands”. Thus, intelligence has long been understood to come together with dexterity, and the lack thereof in artificial systems explains why current robots are mostly limited to pre-programmed tasks in structured environments. In this project named after Aristotle, we aim to develop a new generation of dexterous robotic manipulation systems endowed with a pair of three-fingered hands, human-like vision and tactile sensing, and more importantly, the capacity for never-ending learning and inference to deal with ever more complex manipulation tasks. To achieve this goal, the Aristotle project requires the following components of increasing level of complexity: representation learning for tactile-driven finger control for in-hand manipulation, visuotactile hand-eye calibration fusing tactile and visual sensing modalities, dual-arm coordination that combines representations from the left and right hand, and finally, never-ending lifelong learning to capitalize on the previously learned manipulation skills. As such, the Aristotle project challenges the current 2nd-wave AI paradigm centered on learning from static human-curated datasets, and asserts itself as a 3rd-wave AI initiative by targeting human-level robotic dexterity. The Aristotle project stands out from traditional research proposals thanks to a deep and harmonious 40-year-long partnership between Ecole Centrale de Lyon (ECL) and Technische Universität Darmstadt (TUDa), conducive to forming a joint French-German AI research lab between two world-class AI research teams with complementary expertise in computer vision and robotics, thereby establishing an open platform for collaboration between other AI faculty at both institutions and further European partners. For this purpose, the Aristotle project proposes two co-supervised PhD theses and an annual 4-day open workshop, fostering scientific exchange and Europe-wide collaboration among AI researchers.
Congratulations to my PhD student, Amaury Depierre, for the acceptance of his submission to #ICRA2021 entitled "Optimizing correlated graspability score and grasp regression for better grasp prediction". The video showing the work is here https://youtu.be/KqsIRPL4r1Y; The preprint can be found here: https://arxiv.org/pdf/2002.00872.pdf
In this paper, we extend a state-of-the-art neural network with a scorer which evaluates the graspability of a given position and introduce a novel loss function which correlates regression of grasp parameters with graspability score. We show that this novel architecture improves the performance from 81.95% for a state-of-the-art grasp detection network to 85.74% on Jacquard dataset. Because real-life applications generally feature scenes of multiple objects laid on a variable decor, we also introduce Jacquard+, a test-only extension of Jacquard dataset. Its role is to complete the traditional real robot evaluation by benchmarking the adaptability of a learned grasp prediction model on a different data distribution than the training one while remaining in totally reproducible conditions. Using this novel benchmark and evaluated through the Simulated Grasp Trial criterion, our proposed model outperforms a state-of-the-art one by 7 points.
a 3D GAN for Improved Large_pose facial recognition
Very pleased that our paper on a 3D GAN for Improved Large_pose facial recognition has been accepted in CVPR 2021. An exemplary cooperation between IDEMIA and ECL Liris ! In this work we incorporate a 3D morphable model into the generator of a GAN in order to learn a nonlinear texture model from in-the-wild images. This allows generation
of new, synthetic identities, and manipulation of pose and expression without compromising the identity. Our synthesised data is used to augment training of facial recognition networks with performance evaluated on the challenging CFPW and Cross-Pose LFW datasets. You can find all the details here: https://arxiv.org/pdf/2012.10545.pdf
Annual (virtual) meeting of the Alegoria project which aims at valorizing digitized geographic iconographic heritage, in partnership with IGN, French National Archives, the Nicéphore Niépce Museum, Lavue. Find the whole story: http://alegoria.ign.fr/
The announcement: I'm pleased to invite you to the defence of my PhD thesis, which is entitled: "Data-augmentation with Synthetic Identities for Robust Facial Recognition ".
The defence will take place on Monday 14th December at 3:00 pm (this afternoon!) by videoconference that you be able to watch at the following link:
https://www.youtube.com/channel/UCP12DAKq0l9R35cISaiGmEg/live
The jury is composed of:
Liming CHEN, Director of the thesis, Professor, ECL
Sami ROMDHANI, Supervisor, Doctor, IDEMIA
Stéphane GENTRIC, Supervisor, Doctor, IDEMIA
Dimitris SAMARAS, Rapporteur, Professor, Stony Brook University
Boulbaba BEN AMOR, Rapporteur, Professor, Université de Lille
Bernadette DORIZZI, Examiner, Professor, Institut Télécom, Télécom SudParis
Ioannis KAKADIARIS, Examiner : Professor, University of Houston
Abstract (EN):
In 2014, use of deep neural networks (DNNs) revolutionised facial recognition (FR). DNNs are capable of learning to extract feature-based representations from images that are discriminative and robust to extraneous detail. Arguably, one of the most important factors now limiting the performance of FR algorithms is the data used to train them. High-quality image datasets that are representative of real-world test conditions can be difficult to collect. One potential solution is to augment datasets with synthetic images. This option recently became increasingly viable following the development of generative adversarial networks (GANs) which allow generation of highly realistic, synthetic data samples. This thesis investigates the use of GANs for augmentation of FR datasets. It looks at the ability of GANs to generate new identities, and their ability to disentangle identity from other forms of variation in images. Ultimately, a GAN integrating a 3D model is proposed in order to fully disentangle pose from identity. Images synthesised using the 3D GAN are shown to improve large-pose FR and a state-of-the-art accuracy is demonstrated for the challenging Cross-Pose LFW evaluation dataset.
The final chapter of the thesis evaluates one of the more nefarious uses of synthetic images: the face-morphing attack. Such attacks exploit imprecision in FR systems by manipulating images such that they might be falsely verified as belonging to more than one person. An evaluation of GAN-based face-morphing attacks is provided. Also introduced is a novel, GAN-based morphing method that minimises the distance of the morphed image from the original identities in a biometric feature-space. A potential counter measure to such morphing attacks is to train FR networks using additional, synthetic identities. In this vein, the effect of training using synthetic, 3D GAN data on the success of simulated face-morphing attacks is evaluated.
While robots in various forms are increasingly penetrating our daily life, e.g., automation of repetitive and penible tasks in industry, their use is mostly limited to specific tasks within known environment because of the complexity of the design of robots’ controllers even for very simple operations for humans, e.g., pouring a cup of tea. The CHIRON project aims to develop an AI empowered general purpose robotic system for dexterous manipulation of complex objects in unknown environments. This will be achieved through intuitive embodied robotic teleoperation for shared-controlled tele-manipulation tasks. Although such a system could be used in many applications involving complex manipulation of unknown objects in unstructured or hasardeuse environments, e.g., disasters, space or underwater exploration, the privileged use case within the CHIRON project is assistance for manipulation tasks, e.g., fetching a bottle of water, pouring it into a glass, bringing a new magazine from a shelf, to help “stick-to-bed” patients or elders in their daily life, through an intuitive and embodied robot tele-operated by themself.
The CHIRON project gathers three partners with world class expertise in their complementary AI domains. Prof. Hasegawa’s group brings its unique expertise in the design and embodiment of an extra robotic thumb and intelligent cane for elderly. His challenge within the project is to go much further and design a simple and intuitive VR-based interface to enable a non expert operator to seamlessly manipulate objects through the dual-arm robot and visual and tactile feedback and achieve robotic embodiment without destruction. Prof. Peters’ team is providing their well known rich experience in robotc manipulation, including tactile sensing, robotic tele-operation, learning by demonstration, reinforcement learning. Their challenge in the project is to design a smart dual-arm robot controller sharing the control with the tele-operator however without explicit models on the object, its environment, grippers with tactile sensors and the dual-arm robot. Prof. Chen’s group is bringing their confirmed expertise in computer vision and machine learning for deep understanding of the scene for object manipulation. Their challenge is to make effective object segmentation with their pose and grasp position only with very few data for learning. For this purpose they are investigating few shot learning from the perspective of meta-learning.
In collaboration with EXCOFFIER FRERES and SILEANE, two leading French companies in waste recycling and industrial robotics, I am very pleased to announce that we are going to work on AI enhanced robotised waste sorting for better environment protection within the FAIR WASTES project, a French government supported project worth of 3.73 M€ for the full investment. My group will be involved in the design of AI algorithms, e.g., object instance segmentation and recognition, recognition of materials, AI enhanced grasping. My group at ECL Liris is funded 377 K€ within the project.
Very pleased to annonce that our last work on instance segmentation and layout prediction from a single RGB image will soon appear in International Journal of Computer Vision (IJCV). It was a research work by our former PhD student, Matthieu Grard through an industrial thesis in collaboration with Siléane, a Saint-Etienne-based SME specialised in industrial robotics. In this work, we have proposed a multicameral design of deep auto-encoder composed of subtask-specific lightweight decoder and encoder-decoder units. For further details, please read the preprint here: https://lnkd.in/f2YMupMg
The workshop is taking place on June 4th, 2019 at CNAM within the framework of the open lab GDR ISIS. Please find the full program here.
This book constitutes the refereed proceedings of the 7th International Workshop on Representations, Analysis and Recognition of Shape and Motion from Imaging Data, RFMI 2017, held in Savoi, France, in December 2017.
Find more here: https://www.springer.com/us/book/9783030198152
The objective of this workshop is to present and discuss the latest and most significant trends in the analysis, structuring, and understanding of multimedia contents dedicated to the valorization of heritage, with the emphasis on the unlocking of and access to the big data of the past.
for further details, visit https://sumac2019.ec-lyon.fr
I have given a talk on "giving eyes and intelligence to grasping robots" and attended to his group weekly meeting
I gave a seminar on GAN and Its Applications at Meetup and Lyon Data Science on 12-Mar-2019...https://www.meetup.com/fr-FR/Lyon-Data-Science/events/259498325
Une percée majeure en machine learning est l'apparition récente des réseaux antagonistes génératifs ou en anglais Generatif Adversarial Networks (GANs), qui permet de simuler des données , e.g., visages, pratiquement impossibles de différencier de vraies données. Ses applications sont aussi nombreuses que diversifiées, allant de l'édition de photos jusqu'à la traduction automatique en passant par le transfert de style. Dans cet exposé, Prof. Chen va introduire les principes de base des GANs , en décrire quelques applications , et donner un aperçu de ses derniers travaux, notamment de l'algorithme IVI-GAN, qui permet d'isoler des facteurs de variation dans la génération de données.
Dr. Liming Chen est Professeur à l'Ecole Centrale de Lyon où il mène un groupe de recherche sur la vision par ordinateur, le machine learning et la robotique depuis de nombreuses années. Son groupe a été lauréat des médailles d'or et d'argent du challenge ImageClef sur l'annotation automatique d'images en 2011 et de médaille d'or au challenge Shrec 3D face recognition en 2012. Leurs travaux de recherche ont trouvé de nombreuses applications, e.g., Morphoway, en biométrie pour le portail de passage automatique à l'aéroport, ou encore en bras de manipulation robotique, en partenariat avec des acteurs industriels majeurs.
Références :
- Generative Adversarial Nets, Goodfellow : https://papers.nips.cc/paper/5423-generative-adversarial-nets.pdf
- Conditional Generative Adversarial Nets, Mirza : https://arxiv.org/pdf/1411.1784.pdf
- Intra-class Variation Isolation in Conditional GANs, Chen : https://arxiv.org/pdf/1811.11296.pdf
The second workshop on Face and Gesture Analysis for Health Informatics (FGAHI) will be held in conjunction with CVPR 2019, June 16th - June 21st, Long Beach, CA. The workshop aims to discuss the strengths and major challenges in using computer vision and machine learning of automatic face and gesture analysis for clinical research and healthcare applications. We invite scientists working in related areas of computer vision and machine learning for face and gesture analysis, affective computing, human behavior sensing, and cognitive behavior to share their expertise and achievements in the emerging field of computer vision and machine learning based face and gesture analysis for health informatics.
Full information available here: http://fgahi2019.isir.upmc.fr/
Weaving Ubiquitous Sensing and Computing into Ubiquitous Intelligence
Ubiquitous sensors, devices, networks and information are paving the way towards a smart world in which computational intelligence is distributed throughout the physical environment to provide reliable and relevant services to people. This ubiquitous intelligence will change the computing landscape
because it will enable new breeds of applications and systems to be developed and the realm of computing possibilities will be significantly extended. By enhancing everyday objects with intelligence, many tasks and processes could be simplified, more efficient and more enjoyable. Ubiquitous computing is to create such intelligent/smart environments, services and applications.
The 16th IEEE International Conference on Ubiquitous Intelligence and Computing (UIC 2019) will
include a highly selective program of technical papers, accompanied by workshops, demos, panel
discussions and keynote speeches. We welcome high quality papers that describe original and
unpublished research advancing the state of the art in ubiquitous intelligence and computing.
With Dr.Emmanuel Dellandréa, we are giving a conference on Deep learning for Detecting and Segmenting Objects in the cycle Vision, Image and AI within the Open University of Lyon. 1.
09-APR-2019, at 5-7 pm, Amphithéâtre (rdc), ENSSIB, 17-21 Boulevard du 11 Novembre 1918, 69100 Villeurbanne
In this survey, we propose to discuss knowledge transfer works for vision recognition tasks, and to explore the common rules behind these works. As the primary goal of knowledge transfer methods is to harness learned knowledge for re-use in novel learning tasks, we overview in this survey knowledge transfer methods from the viewpoint of the knowledge being transferred. Specifically, We will firstly discuss the reusable knowledge in a vision recognition task, then we will categorize different kinds of knowledge transfer approaches by where the knowledge comes from and where the knowledge goes to. We aim at finding general rules across different Transfer Learning (TL) settings instead of focusing on their particularities. This viewpoint is in clear contrast to previous surveys on TL. However, any existing method has its own scope of applicability, and we will indicate the applicable scenarios when introducing each method.
See the details here: https://hal.archives-ouvertes.fr/hal-02101005
The central focus of this journal is the computer analysis of pictorial information. Computer Vision and Image Understanding publishes papers covering all aspects of image analysis from the low-level, iconic processes of early vision to the high-level, symbolic processes of recognition and interpretation. A wide range of topics in the image understanding area is covered, including papers offering insights that differ from predominant views. CiteScore:8.7; IF:3.121.
This is made possible thanks to our latest research work on IVI-GAN. Find the full story here: Intra-class Variation Isolation in Conditional GAN
A novel project LEARN-REAL has been recommended for funding within the framework of CHIST-ERA program. The project aims at improving reproducibility in LEARNing physical manipulation skills with simulators using REAListic variations and partners my group at LIRIS ECL with Dr.Sylvain Calinon's group and Dr.André Anjos' group at Idiap Research Institute (Swiss) and Prof.Darwin Caldwell's group at Italian Institute of Technology (IIT, Italy). The project was submitted to the CHIST-ERA 2017 call on Object recognition and manipulation by robots: Data sharing and experiment reproducibility (ORMR). 7 projects out of 26 proposals have been proposed for funding. The share of the project for LIRIS ECL is 250 K€. The project is expected to start on 01-Apr-2019.
A novel special issue on face analysis for applications for ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM) has just been launched for a submission deadline set to Oct.14, 2018. I am serving as guest editor along with Prof.Pietro Pala (University of Florence, Italy), Ass. Prof.Di Huang (Beihang University, China), Ass.Prof.Xiaoming Liu (Michigan State University, USA), and Ass.Prof.Stefanos Zafeiriou (Imperial College, London, UK). For full details, please visit the webpage here: https://docs.google.com/document/d/1m1qLbzY7p7YLEP9ic4vZ65CiWJAzNU7EuS0KMn2YvZ8/edit?usp=sharing
The 14th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2019) will be held in Lille, France, 14 -18 May 2019. I am serving as area chair of this fascinating international conference covering advances in fundamental computer vision, pattern ecognition and computer graphics; machine learning techniques relevant to face, gesture, and body motion;new algorithms and applications...Visit the website http://www.fg2019.org further details...
The 4D Vision project, that I have been coordinating with Prof.Dimitris Samaras at New York State University at Stony Brook, has just been notified for its second year support. The 4D Vision project is a Franco-american research project funded by the Partner University Fund (PUF) for a total cost up to $1 240 337 of which $150 000 are granted by PUF for a 36 month duration. It focuses on cutting edge research issues on 4D image data and gathers Prof.Dimitris Samaras and Prof.David Xianfeng GU at New York State University Stony Brook in partnership with research groups from Ecole Centrale de Paris (Prof. Nikos Paragios) and University of Houston (Prof.Ioannis Kakadiaris). Visit the project's website here for more details.
I co-organized with Prof.Boulbaba Ben Amor at IMT Lille Douai/CRIStAL (UMR CNRS 9189) the 7th International Workshop on Representation, analysis and recognition of shape and motion from Image Data (RFMI 2017) on December 17-20, 2017, at Centre Paul-Langevin, Aussois, France. The newsletter of the International Association for Pattern Recognition (IAPR) at its issue of April 2018 talks about this event. Find more here.
A novel article on Children facial expression production Children facial expression production: influence of age, gender, emotion subtype, elicitation condition and culture is published in Frontiers in Psychology, section Emotion Science. Il results from a collaboration with psychologists and doctors within the the ANR supported Jemime project whose goal is to develop the serious game in to help children with autism spectrum disorder (ASD) gain emotional intelligence, i.e., to learn to mimic facial and vocal emotions and to express the proper emotion depending on the context. My colleague Prof.David Cohen has given an interview on the meaning of this research at France Culture here: https://www.franceculture.fr/emissions/matieres-a-penser-avec-serge-tisseron/autisme-jeux-serieux-et-robotique-realite-tangible-ou-abus-de-langage
We are releasing a large scale dataset for robotic grasping detection: the Jacquard grasping Dataset ! Built upon ShapeNet, it features at a scale of millions varied object grasping positions for a large diversity of more than 11k objects. Beyond the simulation of the grasping scene with the underlying object, the ground truth for a successful grasping position is also based on tentatives of a simulated grasping robot. Fore more details, please read our archived paper here: https://arxiv.org/abs/1803.11469
I am co-organizing a one day workshop entitled "adding semantic information for computer vision" within the framework of the GDR open lab ISIS on April 6, 2018 at the amphitheater Abbé Prouvé, CNAM (292 rue Saint-Martin, 75003 Paris). The program is now available here: http://gdr-isis.fr/index.php?page=reunion&idreunion=358
The paper "Visual and Semantic Knowledge Transfer for Large Scale Semi-supervised Object Detection" , appeared in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), DOI: 10.1109/TPAMI.2017.2771779 on 3-Nov-2017, will be presented as poster at the ARC6 workshop on Deep learning & reinforcement learning on 30-Mar-2018. The workshop program can be found here: http://registration.listic.org/dldrl/
A joint work on "Accurate Facial Parts Localization and Deep Learning for 3D Facial Expression Recognition" with my colleague, Dr.Hongying Meng at UCL, UK and Dr.Huibin Li at XJTU, China, has been accepted by FG'2018. It is also available here: http://arxiv.org/abs/1803.05846.
Our work on "Improving Shadow Suppression for illumination Robust Face Recognition" was accepted in IEEE TPAMI on 09-Jan-2018 and is online: https://doi.org/10.1109/TPAMI.2018.2803179 . It was a major research topic of Zhang Wuming, my former PhD student;
We released a novel unsupervised domain adaptation method at ArXiv, namely Discriminative and Geometry Aware Unsupervised Domain Adaptation; In this method, we show that domain adaptation should be discriminative and take into account the underlying data geometric structure ;
With my former postdoc, Dr. Md Abul Hasnat and colleagues at Morpho, we released at ArXiv a novel deep learning scheme based on von Mises-Fisher mexture model and apply it to face recognition,. We show the proposed deep learning scheme demonstrates outstanding generalization skills on a number of standard face benchmarks, including LFW, CACD.