A5017977139- Face recognition and analysis
- Face and Expression Recognition
- 3D Shape Modeling and Analysis
- Image Retrieval and Classification Techniques
- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- Image Processing and 3D Reconstruction
- Emotion and Mood Recognition
- Morphological variations and asymmetry
- Hand Gesture Recognition Systems
- Human Motion and Animation
- Computer Graphics and Visualization Techniques
- Gait Recognition and Analysis
- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Robotics and Sensor-Based Localization
- Video Analysis and Summarization
- Generative Adversarial Networks and Image Synthesis
- Digital Image Processing Techniques
- Anomaly Detection Techniques and Applications
- Neural Networks and Applications
- Advanced Numerical Analysis Techniques
- 3D Surveying and Cultural Heritage
- Machine Learning and Data Classification
- EEG and Brain-Computer Interfaces
IMT Nord Europe
2017-2025
Université de Lille
2016-2025
Centre de Recherche en Informatique
2016-2025
Centre National de la Recherche Scientifique
2015-2024
Institut Mines-Télécom
2011-2024
École Centrale de Lille
2021-2024
Chouaib Doukkali University
2022-2024
Mohammed V University
2023
Centre de Recherche en Informatique, Signal et Automatique de Lille
2015-2022
King Abdulaziz University
2021
Recognizing human actions in 3-D video sequences is an important open problem that currently at the heart of many research domains including surveillance, natural interfaces and rehabilitation. However, design development models for action recognition are both accurate efficient a challenging task due to variability pose, clothing appearance. In this paper, we propose new framework extract compact representation captured through depth sensor, enable recognition. The proposed solution...
We propose a novel geometric framework for analyzing 3D faces, with the specific goals of comparing, matching, and averaging their shapes. Here we represent facial surfaces by radial curves emanating from nose tips use elastic shape analysis these to develop Riemannian shapes full surfaces. This representation, along metric, seems natural measuring deformations is robust challenges such as large expressions (especially those open mouths), pose variations, missing parts, partial occlusions...
In this paper, we propose a method for three-dimensional (3D)-model indexing based on two-dimensional (2D) views, which call adaptive views clustering (AVC). The goal of is to provide an "optimal" selection 2D from 3D model, and probabilistic Bayesian 3D-model retrieval these views. characteristic view algorithm uses statistical model distribution scores select the optimal number Starting fact that all do not have equal importance, also introduce novel approach improve retrieval. Finally,...
We study shapes of facial surfaces for the purpose face recognition. The main idea is to 1) represent by unions level curves, called depth function and 2) compare implicitly using curves. latter performed a differential geometric approach that computes geodesic lengths between closed curves on shape manifold. These ideas are demonstrated nearest-neighbor classifier two 3D databases: Florida State University Notre Dame, highlighting good recognition performance.
In this paper, the problem of person-independent facial expression recognition is addressed on 3D shapes. To end, an original approach proposed that computes SIFT descriptors a set landmarks depth images, and then selects subset most relevant features. Using SVM classification selected features, average rate 77.5% BU-3DFE database has been obtained. Comparative evaluation common experimental setup, shows our solution able to obtain state art results.
Non-rigid 3D shape retrieval has become an important research topic in content-based object retrieval. The aim of this track is to measure and compare the performance non-rigid methods implemented by different participants around world. based on a new benchmark, which contains 600 watertight triangle meshes that are equally classified into 30 categories. In track, 25 runs have been submitted 9 groups their accuracies were evaluated using 6 commonly-utilized measures.
Recently, wearable emotion recognition based on peripheral physiological signals has drawn massive attention due to its less invasive nature and applicability in real-life scenarios. However, how effectively fuse multimodal data remains a challenging problem. Moreover, traditional fully-supervised approaches suffer from overfitting given limited labeled data. To address the above issues, we propose novel self-supervised learning (SSL) framework for recognition, where efficient fusion is...
Abstract This paper presents a novel approach for fast and efficient partial shape retrieval on collection of 3D shapes. Each is represented by Reeb graph associated with geometrical signatures. Partial similarity between two shapes evaluated computing variant their maximum common sub‐graph. By investigating theory, we take advantage its intrinsic properties at levels. First, show that the segmentation provides charts disk or annulus topology only. control enables computation concise...
We utilize ideas from two growing but disparate in computer vision-shape analysis using tools differential geometry and feature selection machine learning-to select highlight salient geometrical facial features that contribute most 3-D face recognition gender classification. First, a large set of geometries curve are extracted level sets (circular curves) streamlines (radial the Euclidean distance functions surface; together they approximate surfaces with arbitrarily high accuracy. Then, we...
In this paper, we present an automatic approach for facial expression recognition from 3-D video sequences. the proposed solution, faces are represented by collections of radial curves and a Riemannian shape analysis is applied to effectively quantify deformations induced expressions in given subsequence frames. This obtained dense scalar field, which denotes shooting directions geodesic paths constructed between pairs corresponding two faces. As resulting fields show high dimensionality,...
In this paper, we propose a novel space-time geometric representation of human landmark configurations and derive tools for comparison classification. We model the temporal evolution landmarks as parametrized trajectories on Riemannian manifold positive semidefinite matrices fixed-rank. Our has benefit to bring naturally second desirable quantity when comparing shapes-the spatial covariance-in addition conventional affine-shape representation. derived then computational rate-invariant...
In this work, we propose a novel approach for generating videos of the six basic facial expressions given neutral face image. We to exploit geometry by modeling landmarks motion as curves encoded points on hypersphere. By proposing conditional version manifold-valued Wasserstein generative adversarial network (GAN) generation hypersphere, learn distribution expression dynamics different classes, from which synthesize new motions. The resulting motions can be transformed sequences and then...
This work is aimed at the detection of adult images that appear in Internet. Skin paramount importance images. We build a maximum entropy model for this task. model, called First Order Model paper, subject to constraints on color gradients neighboring pixels. Parameter estimation as well optimization cannot be tackled without approximations. With Bethe tree approximation, parameter eradicated and Belief Propagation algorithm permits obtain exact fast solution skin probabilities pixel...
Abstract This paper presents a 3D‐mesh segmentation algorithm based on learning approach. A large database of manually segmented 3D‐meshes is used to learn boundary edge function. The function learned using classifier which automatically selects from pool geometric features the most relevant ones detect candidate edges. We propose processing pipeline that produces smooth closed boundaries this successively set contours, closes them and optimizes snake movement. Our was evaluated...
In this paper, we propose an holistic, fully automatic approach to 3D Facial Expression Recognition (FER). A novel facial representation, namely Differential Mean Curvature Maps (DMCMs), is proposed capture both global and local surface deformations which typically occur during expressions. These DMCMs are directly extracted from depth images, by calculating the mean curvatures thanks integral computation. To account for morphology variations, they further normalized through aspect ratio...
Recent breakthroughs in deep learning using automated measurement of face and head motion have made possible the first objective depression severity. While powerful, approaches lack interpretability. We developed an interpretable method automatically measuring severity that uses barycentric coordinates facial landmarks a Lie-algebra based rotation matrix 3D motion. Using these representations, kinematic features are extracted, preprocessed, encoded Gaussian Mixture Models (GMM) Fisher vector...