Saida Brondel - Profile on Academia.edu (original) (raw)

Papers by Saida Brondel

アニメーション化された3D骨格からの身体表情認識【Powered by NICT】

IEEE Conference Proceedings, 2016

HAL (Le Centre pour la Communication Scientifique Directe), Oct 13, 2020

Although predictive forward gait simulators have existed for a little while, only few sensitivity... more Although predictive forward gait simulators have existed for a little while, only few sensitivity studies have been proposed. We present a study analysing precision, accuracy, stability and sensitivity of such simulator. We studied a gait simulator based on the trajectory mimicking method proposed by Lee et al. . We assessed the effect of motion variation on the resulting gait simulation using the sensitivity analysis method proposed by Saltelli et al. . Our method supports the use of this simulator for prediction.

Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, 2017

The detection of moving objects in the video stream of a moving camera is a complex task. Static ... more The detection of moving objects in the video stream of a moving camera is a complex task. Static objects appear moving in the video stream as moving objects. Thus, it is difficult to identify motions that belong to moving objects because they are hidden by those of static objects. To detect moving objects we propose a novel geometric constraint based on 2D and 3D information. A sparse reconstruction of the visible part of the scene is performed in order to detect motions in the 3D space where the scene perception is not deformed by the camera motion. A first labeling estimation is performed in the 3D space and then apparent motions in the video stream of the moving camera are used to validate the estimation. Labels are computed from confidence values which are updated at each frame according to the geometric constraint. Our method can detect several moving objects in complex scenes with high parallax.

Separating superfluous from essential: which facial region (s) holds the key for expression recognition?

Abstract Communication is the key aspect of everyday life and facial expressions provide the most... more Abstract Communication is the key aspect of everyday life and facial expressions provide the most effective way of non-verbal relations. Human-Computer interactions, social robots, deceit detection and behavior monitoring are some of the potential application areas that ...

IEEE Conference Proceedings, 2016

We present a novel and generic framework for the recognition of body expressions using human post... more We present a novel and generic framework for the recognition of body expressions using human postures. Motivated by the state of the art from the domain of psychology, our approach recognizes expression by analyzing sequence of pose. Features proposed in this article are computationally simple and intuitive to understand. They are based on visual cues and provide in-depth understanding of body postures required to recognize body expressions. We have evaluated our approach on different databases with heterogeneous movements and body expressions. Our recognition results exceeds state of the art for some database and for others we obtain results at par with state of the art.

Virtual arm for the phantom pain therapy

HAL (Le Centre pour la Communication Scientifique Directe), Nov 16, 2005

Phantom limb pain appears where an amputated limb used to be and is often excruciating and almost... more Phantom limb pain appears where an amputated limb used to be and is often excruciating and almost impossible to treat. Recently, two experiments on hand amputees in the neuroscience area have shown that if a patient has the visual illusion of his phantom hand in a way that he is able to move it, the phantom pain can decrease. In this context, we propose an augmented reality system. Our interactive system provides to single arm amputee the image of himself augmented by a virtual arm. It is based on the valid arm detection on each frame from a single camera, in order to use it to create on the image the virtual arm on the amputated side. By giving a immersive sensation and a realistic image of his complete body to the subject, we anticipate that this could lead to a sense of re-embodiment in the virtual limb as with previous system dedicated to the hand. And then, lead to a reduction of the pain. Moreover, by allowing more freedom we could also expect a better impact rate. Later, we plan to extend it to other limb.

We present a novel and generic framework for the recognition of body expressions using human post... more We present a novel and generic framework for the recognition of body expressions using human postures. Motivated by the state of the art from the domain of psychology, our approach recognizes expression by analyzing sequence of pose. Features proposed in this article are computationally simple and intuitive to understand. They are based on visual cues and provide in-depth understanding of body postures required to recognize body expressions. We have evaluated our approach on different databases with heterogeneous movements and body expressions. Our recognition results exceeds state of the art for some database and for others we obtain results at par with state of the art.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 10, 2021

Accurate and specific predictive numerical simulation of the human gait can help in many clinical... more Accurate and specific predictive numerical simulation of the human gait can help in many clinical routines. Predictive methods often track modifications applied to a reference motion but that usually do not take into account the characteristics and the stability of the predicted motion. We propose an approach that includes a cost function, computed on the predicted motion, which optimization preserves these concerns. Experimental studies on subjects with different walking patterns confirmed that our method preserves the characteristics of the gait.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 24, 2021

Performing a subject specific and accurate predictive numerical gait simulation can be of great h... more Performing a subject specific and accurate predictive numerical gait simulation can be of great help in many clinical tasks. Though predictive methods often take into account the modifications applied to a reference motion, they are not always able to include the characteristics and the stability of the predicted motion. We propose an optimization-based approach that includes the resulting characteristics of the predicted motion. The optimization is enhanced by the use of parametric curves to represent the motion trajectories. Experimental studies on subjects with different gait patterns confirmed that our method preserves the characteristics of the gait. CCS Concepts •Computing methodologies → Physical simulation; Interactive simulation; Optimization algorithms; Control methods; •Applied computing → Health informatics; 42 open-loop angular target and an adaptive correction. The open-loop 43 angular target is evaluated from the kinematics of one reference gait 44 cycle. The adaptive correction is computed by the neural network 45 from the current pose of the character and the current percent of 46 gait cycle denoted as φ. 47 an objective function composed of a weighted combination of two 73 terms. The first term estimates the relevancy of the produced mo-74 tion by penalizing simulations for which the virtual character falls 75 or collisions occur between the legs during the first 15 gait cycles. 76 The second term depends on the targeted gait characteristics. In our 77 example, the stiff-knee gait, it penalizes the knee flexion, measured 78 over the last 10 gait cycles. 79 Each simulation takes about 1s to execute (20 times faster than 80 real-time). It was important to use a method that converge with 81 a minimum number of evaluation and without the need to eval-82 uate gradients because the problem is discontinuous. We chose 83 the Covariance Matrix Adaptation Evolution Strategies (CMA-ES) 84 method as our optimization process [Han07]. 85 With the discrete representation of the motion, there are more 86 than 1000 parameters to optimize. CMA-ES shows best perfor-87 mance with less than 100 parameters so two strategies were used 88 to reduce the search space. First, we compute a parametric approx-89 imation of each joint trajectory of the kinematics data, allowing us 90 to model a full trajectory from few control points only. Then, a vi-91 sual comparison between trajectories from both conditions C0 and 92 C1 and knowledge from gait analysis of the targeted pattern is used 93 to identify a subset of the trajectories to include in the optimization. 94 2.3. Parametric trajectories representation 95 We were looking for a parametric description of the trajectories 96 with the following features : accurate approximation with a small 97 number of parameters, C 2 continuity and fast evaluation. Non-98 Uniform Rational Basis Spline (NURBS) presents theses advan-99 tages. We choose to use cubic periodic NURBS for all trajectories 100 except for the transverse plan pelvis coordinates. For those coor-101 dinates we chose cubic B-splines. The optimum placement of the 157 mum variation of 2% from the one used during the optimization 158 process. 159 On Figure 1, we observe that the simulated kinematics has 160 been affected by the modifications of the reference motion. Dis-161 persion between simulated kinematics increased and is more pro-162 nounced during the first half of the swing phase of the right leg 163 (φ < 25%). The mean value of the simulated kinematics has signif-164 icantly changed for many degrees of freedom even for the trajecto-165 ries that were not included in the optimisation. The left hip flexion gait using Lissajous figures.

HAL (Le Centre pour la Communication Scientifique Directe), Oct 23, 2017

Rsum Using a lower simulation frequency for physics-based control of virtual characters fre... more Rsum Using a lower simulation frequency for physics-based control of virtual characters frees computation time that can be used for more complex environment. However, using low simulation frequency may introduce instabilities inside the simulation. In this paper, we demonstrate that even a simple control strategy can be used at a low simulation frequency by adapting the control parameters. Indeed we show that lower frequencies hold a more restrictive space of possible control parameters than higher ones. We propose a method to find optimized control parameters for frequencies as low as 200Hz. As using such low frequencies may introduce foot-ground contact instabilities, we also introduce an additional control feedback on the stance leg. Our controller shows similar robustness as high frequency controller while using 0.8ms per simulation step.

The Use of Constraints for Calibration-Free 3D Metric Reconstruction: From Theory to Applications

Series in computer vision, Sep 1, 2011

The use of constraints for calibration-free 3D metric reconstruction: from theory to applications.

Computer Animation and Virtual Worlds, Jun 1, 2021

Figure 1: Sample results of our method in various configuration. Our method is able to generate p... more Figure 1: Sample results of our method in various configuration. Our method is able to generate plausible poses given a starting pose (on the left) and some targets (in red), respecting skeleton constraints without having to explicitly specify them.

Springer eBooks, Oct 2, 2021

Although predictive forward gait simulators have existed for a little while, only few sensitivity... more Although predictive forward gait simulators have existed for a little while, only few sensitivity studies have been proposed. We present a study analysing precision, accuracy, stability and sensitivity of such simulator. We studied a gait simulator based on the trajectory mimicking method proposed by Lee et al. [1]. We assessed the effect of motion variation on the resulting gait simulation using the sensitivity analysis method proposed by Saltelli et al. [2]. Our method supports the use of this simulator for prediction.

Figure 1: We propose a physics-based controller able to simulate the walk of partially immersed c... more Figure 1: We propose a physics-based controller able to simulate the walk of partially immersed characters. Our controller provides robust balancing, gait style adaptation to the environment and precise speed tracking. In this figure, we observe that the swing foot gets near the water surface before the character starts moving forward.

HAL (Le Centre pour la Communication Scientifique Directe), Feb 10, 2023

In animation, style can be considered as a distinctive layer over the content of a motion, allowi... more In animation, style can be considered as a distinctive layer over the content of a motion, allowing a character to achieve the same gesture in various ways. Editing existing animation to modify the style while keeping the same content is an interesting task, which can facilitate the re-use of animation data and cut down on production time. Existing animation edition methods either work directly on the motion data, providing precise but tedious tools, or manipulate semantic style categories, taking control away from the user. As a middle ground, we propose a new character motion edition paradigm allowing higher-level manipulations without sacrificing controllability. We describe the concept of pose metrics, objective value functions which can be used to edit animation, leaving the style interpretation up to the user. We then propose an edition pipeline to edit animation data using pose metrics.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 20, 2017

IEEE Transactions on Affective Computing, Dec 2, 2022

Many areas in computer science are facing the need to analyze, quantify and reproduce movements e... more Many areas in computer science are facing the need to analyze, quantify and reproduce movements expressing emotions. This paper presents a systematic review of the intelligible factors involved in the expression of emotions in human movement and posture. We have gathered the works that have studied and tried to identify these factors by sweeping many disciplinary fields such as psychology, biomechanics, choreography, robotics and computer vision. These researches have each used their own definitions, units and emotions, which prevents a global and coherent vision. We propose a meta-analysis approach that cross-references and aggregates these researches in order to have a unified list of expressive factors quantified for each emotion. A calculation method is then proposed for each of the expressive factors and we extract them from an emotionally annotated animation dataset: Emilya. The comparison between the results of the meta-analysis and the Emilya analysis reveals high correlation rates, which validates the relevance of the quantified values obtained by both methodologies. The analysis of the results raises interesting perspectives for future research in affective computing.

Optimizations for predictive–corrective particle-based fluid simulation on GPU

The Visual Computer, 2022

The use of particles-based simulations to produce fluid animations is nowadays a frequently used ... more The use of particles-based simulations to produce fluid animations is nowadays a frequently used method by both the industrial and research sectors. Although there are many variations of the smoothed particle hydrodynamics (SPH) algorithm currently being used, they all have the common characteristic of being highly parallel in nature. They are therefore frequently implemented on graphics processing units (GPUs) to benefit of high computation capacities of modern GPUs. However, such optimizations require specific optimizations to make use of the full capacity of the GPU, with sometimes optimizations being contradictory to optimizations used in CPU implementations. In this paper, we explored various optimizations on a GPU implementation of a recent particle-based fluid simulation algorithm using an iterative pressure solver. In particular, we focused on CPU optimizations that have not been thoroughly studied for GPU implementations: the indexing for the neighbor’s structure, the frequency of the sorting of the fluid particles, the use of lookup tables for the kernel function computations and the use of a warm-start to improve the performance of the iterative pressure solver. We show that some of these optimizations are only effective for very specific hardware configurations and sometimes even impact the performance negatively. We also show that the warm-start reduces the computation time but introduces a cyclic instability in the simulation. We propose a solution to reduce this instability without requiring to modify the implementation of the fluid algorithm.

2020 International Joint Conference on Neural Networks (IJCNN), 2020

Automatic Hand detection in vehicles achieved best performances using deep learning approach. MS-... more Automatic Hand detection in vehicles achieved best performances using deep learning approach. MS-FRCNN and MS-RFCN are the models that achieve the state of the art results on the viva hand detection challenge. These models, unfortunately, require important resources to operate (more than 7 Giga bytes GPU) which presents an important limit for their large scale applications. In this paper, we address the viva challenge of hand detection in vehicles using simpler models and low computer components (2 Giga bytes GPU). We found that the ZF model is highly adequate for the problem and more importantly, it overpasses remarkably the precisions of complex model of literature MS- FRCNN. The experiment result of ZF model on the viva database shows (AP = 91.5%, AR = 85.7% at level one and AP = 80.5%, AR = 69.5% and Fps = 0.4 s at level two in measures of Average Precision (AP), Average Recall(AR) and Frame per seconds (Fps). These results of ZF reached the fourth place on the viva hand detecti...

Nous presentons une approche pour la reconnaissance d'expressions corporelles qui accompagnen... more Nous presentons une approche pour la reconnaissance d'expressions corporelles qui accompagnent le mouvement d'un personnage reel. L'idee developpee dans cet article s'inspire de travaux en syntheses d'animations. Notre sys-teme se base sur l'analyse de la difference obtenue dans le domaine frequentiel entre un mouvement expressif et un mouvement neutre. Dans ce papier, nous introduisons la notion de mouvement neutre et proposons un algorithme qui permet de l'obtenir par la minimisation d'une fonction de cout global. Le cout global tient compte de la distance parcourue par les differentes articulations et l'accelera-tion de chaque articulation au cours du mouvement. Nous avons evalue notre methode sur differentes bases de don-nees contenant des mouvements et expressions corporelles heterogenes. Les resultats obtenus sont encourageants et confirment la validite de notre demarche.

アニメーション化された3D骨格からの身体表情認識【Powered by NICT】

IEEE Conference Proceedings, 2016

HAL (Le Centre pour la Communication Scientifique Directe), Oct 13, 2020

Although predictive forward gait simulators have existed for a little while, only few sensitivity... more Although predictive forward gait simulators have existed for a little while, only few sensitivity studies have been proposed. We present a study analysing precision, accuracy, stability and sensitivity of such simulator. We studied a gait simulator based on the trajectory mimicking method proposed by Lee et al. . We assessed the effect of motion variation on the resulting gait simulation using the sensitivity analysis method proposed by Saltelli et al. . Our method supports the use of this simulator for prediction.

Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, 2017

The detection of moving objects in the video stream of a moving camera is a complex task. Static ... more The detection of moving objects in the video stream of a moving camera is a complex task. Static objects appear moving in the video stream as moving objects. Thus, it is difficult to identify motions that belong to moving objects because they are hidden by those of static objects. To detect moving objects we propose a novel geometric constraint based on 2D and 3D information. A sparse reconstruction of the visible part of the scene is performed in order to detect motions in the 3D space where the scene perception is not deformed by the camera motion. A first labeling estimation is performed in the 3D space and then apparent motions in the video stream of the moving camera are used to validate the estimation. Labels are computed from confidence values which are updated at each frame according to the geometric constraint. Our method can detect several moving objects in complex scenes with high parallax.

Separating superfluous from essential: which facial region (s) holds the key for expression recognition?

Abstract Communication is the key aspect of everyday life and facial expressions provide the most... more Abstract Communication is the key aspect of everyday life and facial expressions provide the most effective way of non-verbal relations. Human-Computer interactions, social robots, deceit detection and behavior monitoring are some of the potential application areas that ...

IEEE Conference Proceedings, 2016

We present a novel and generic framework for the recognition of body expressions using human post... more We present a novel and generic framework for the recognition of body expressions using human postures. Motivated by the state of the art from the domain of psychology, our approach recognizes expression by analyzing sequence of pose. Features proposed in this article are computationally simple and intuitive to understand. They are based on visual cues and provide in-depth understanding of body postures required to recognize body expressions. We have evaluated our approach on different databases with heterogeneous movements and body expressions. Our recognition results exceeds state of the art for some database and for others we obtain results at par with state of the art.

Virtual arm for the phantom pain therapy

HAL (Le Centre pour la Communication Scientifique Directe), Nov 16, 2005

Phantom limb pain appears where an amputated limb used to be and is often excruciating and almost... more Phantom limb pain appears where an amputated limb used to be and is often excruciating and almost impossible to treat. Recently, two experiments on hand amputees in the neuroscience area have shown that if a patient has the visual illusion of his phantom hand in a way that he is able to move it, the phantom pain can decrease. In this context, we propose an augmented reality system. Our interactive system provides to single arm amputee the image of himself augmented by a virtual arm. It is based on the valid arm detection on each frame from a single camera, in order to use it to create on the image the virtual arm on the amputated side. By giving a immersive sensation and a realistic image of his complete body to the subject, we anticipate that this could lead to a sense of re-embodiment in the virtual limb as with previous system dedicated to the hand. And then, lead to a reduction of the pain. Moreover, by allowing more freedom we could also expect a better impact rate. Later, we plan to extend it to other limb.

We present a novel and generic framework for the recognition of body expressions using human post... more We present a novel and generic framework for the recognition of body expressions using human postures. Motivated by the state of the art from the domain of psychology, our approach recognizes expression by analyzing sequence of pose. Features proposed in this article are computationally simple and intuitive to understand. They are based on visual cues and provide in-depth understanding of body postures required to recognize body expressions. We have evaluated our approach on different databases with heterogeneous movements and body expressions. Our recognition results exceeds state of the art for some database and for others we obtain results at par with state of the art.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 10, 2021

Accurate and specific predictive numerical simulation of the human gait can help in many clinical... more Accurate and specific predictive numerical simulation of the human gait can help in many clinical routines. Predictive methods often track modifications applied to a reference motion but that usually do not take into account the characteristics and the stability of the predicted motion. We propose an approach that includes a cost function, computed on the predicted motion, which optimization preserves these concerns. Experimental studies on subjects with different walking patterns confirmed that our method preserves the characteristics of the gait.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 24, 2021

Performing a subject specific and accurate predictive numerical gait simulation can be of great h... more Performing a subject specific and accurate predictive numerical gait simulation can be of great help in many clinical tasks. Though predictive methods often take into account the modifications applied to a reference motion, they are not always able to include the characteristics and the stability of the predicted motion. We propose an optimization-based approach that includes the resulting characteristics of the predicted motion. The optimization is enhanced by the use of parametric curves to represent the motion trajectories. Experimental studies on subjects with different gait patterns confirmed that our method preserves the characteristics of the gait. CCS Concepts •Computing methodologies → Physical simulation; Interactive simulation; Optimization algorithms; Control methods; •Applied computing → Health informatics; 42 open-loop angular target and an adaptive correction. The open-loop 43 angular target is evaluated from the kinematics of one reference gait 44 cycle. The adaptive correction is computed by the neural network 45 from the current pose of the character and the current percent of 46 gait cycle denoted as φ. 47 an objective function composed of a weighted combination of two 73 terms. The first term estimates the relevancy of the produced mo-74 tion by penalizing simulations for which the virtual character falls 75 or collisions occur between the legs during the first 15 gait cycles. 76 The second term depends on the targeted gait characteristics. In our 77 example, the stiff-knee gait, it penalizes the knee flexion, measured 78 over the last 10 gait cycles. 79 Each simulation takes about 1s to execute (20 times faster than 80 real-time). It was important to use a method that converge with 81 a minimum number of evaluation and without the need to eval-82 uate gradients because the problem is discontinuous. We chose 83 the Covariance Matrix Adaptation Evolution Strategies (CMA-ES) 84 method as our optimization process [Han07]. 85 With the discrete representation of the motion, there are more 86 than 1000 parameters to optimize. CMA-ES shows best perfor-87 mance with less than 100 parameters so two strategies were used 88 to reduce the search space. First, we compute a parametric approx-89 imation of each joint trajectory of the kinematics data, allowing us 90 to model a full trajectory from few control points only. Then, a vi-91 sual comparison between trajectories from both conditions C0 and 92 C1 and knowledge from gait analysis of the targeted pattern is used 93 to identify a subset of the trajectories to include in the optimization. 94 2.3. Parametric trajectories representation 95 We were looking for a parametric description of the trajectories 96 with the following features : accurate approximation with a small 97 number of parameters, C 2 continuity and fast evaluation. Non-98 Uniform Rational Basis Spline (NURBS) presents theses advan-99 tages. We choose to use cubic periodic NURBS for all trajectories 100 except for the transverse plan pelvis coordinates. For those coor-101 dinates we chose cubic B-splines. The optimum placement of the 157 mum variation of 2% from the one used during the optimization 158 process. 159 On Figure 1, we observe that the simulated kinematics has 160 been affected by the modifications of the reference motion. Dis-161 persion between simulated kinematics increased and is more pro-162 nounced during the first half of the swing phase of the right leg 163 (φ < 25%). The mean value of the simulated kinematics has signif-164 icantly changed for many degrees of freedom even for the trajecto-165 ries that were not included in the optimisation. The left hip flexion gait using Lissajous figures.

HAL (Le Centre pour la Communication Scientifique Directe), Oct 23, 2017

Rsum Using a lower simulation frequency for physics-based control of virtual characters fre... more Rsum Using a lower simulation frequency for physics-based control of virtual characters frees computation time that can be used for more complex environment. However, using low simulation frequency may introduce instabilities inside the simulation. In this paper, we demonstrate that even a simple control strategy can be used at a low simulation frequency by adapting the control parameters. Indeed we show that lower frequencies hold a more restrictive space of possible control parameters than higher ones. We propose a method to find optimized control parameters for frequencies as low as 200Hz. As using such low frequencies may introduce foot-ground contact instabilities, we also introduce an additional control feedback on the stance leg. Our controller shows similar robustness as high frequency controller while using 0.8ms per simulation step.

The Use of Constraints for Calibration-Free 3D Metric Reconstruction: From Theory to Applications

Series in computer vision, Sep 1, 2011

The use of constraints for calibration-free 3D metric reconstruction: from theory to applications.

Computer Animation and Virtual Worlds, Jun 1, 2021

Figure 1: Sample results of our method in various configuration. Our method is able to generate p... more Figure 1: Sample results of our method in various configuration. Our method is able to generate plausible poses given a starting pose (on the left) and some targets (in red), respecting skeleton constraints without having to explicitly specify them.

Springer eBooks, Oct 2, 2021

Although predictive forward gait simulators have existed for a little while, only few sensitivity... more Although predictive forward gait simulators have existed for a little while, only few sensitivity studies have been proposed. We present a study analysing precision, accuracy, stability and sensitivity of such simulator. We studied a gait simulator based on the trajectory mimicking method proposed by Lee et al. [1]. We assessed the effect of motion variation on the resulting gait simulation using the sensitivity analysis method proposed by Saltelli et al. [2]. Our method supports the use of this simulator for prediction.

Figure 1: We propose a physics-based controller able to simulate the walk of partially immersed c... more Figure 1: We propose a physics-based controller able to simulate the walk of partially immersed characters. Our controller provides robust balancing, gait style adaptation to the environment and precise speed tracking. In this figure, we observe that the swing foot gets near the water surface before the character starts moving forward.

HAL (Le Centre pour la Communication Scientifique Directe), Feb 10, 2023

In animation, style can be considered as a distinctive layer over the content of a motion, allowi... more In animation, style can be considered as a distinctive layer over the content of a motion, allowing a character to achieve the same gesture in various ways. Editing existing animation to modify the style while keeping the same content is an interesting task, which can facilitate the re-use of animation data and cut down on production time. Existing animation edition methods either work directly on the motion data, providing precise but tedious tools, or manipulate semantic style categories, taking control away from the user. As a middle ground, we propose a new character motion edition paradigm allowing higher-level manipulations without sacrificing controllability. We describe the concept of pose metrics, objective value functions which can be used to edit animation, leaving the style interpretation up to the user. We then propose an edition pipeline to edit animation data using pose metrics.

HAL (Le Centre pour la Communication Scientifique Directe), Nov 20, 2017

IEEE Transactions on Affective Computing, Dec 2, 2022

Many areas in computer science are facing the need to analyze, quantify and reproduce movements e... more Many areas in computer science are facing the need to analyze, quantify and reproduce movements expressing emotions. This paper presents a systematic review of the intelligible factors involved in the expression of emotions in human movement and posture. We have gathered the works that have studied and tried to identify these factors by sweeping many disciplinary fields such as psychology, biomechanics, choreography, robotics and computer vision. These researches have each used their own definitions, units and emotions, which prevents a global and coherent vision. We propose a meta-analysis approach that cross-references and aggregates these researches in order to have a unified list of expressive factors quantified for each emotion. A calculation method is then proposed for each of the expressive factors and we extract them from an emotionally annotated animation dataset: Emilya. The comparison between the results of the meta-analysis and the Emilya analysis reveals high correlation rates, which validates the relevance of the quantified values obtained by both methodologies. The analysis of the results raises interesting perspectives for future research in affective computing.

Optimizations for predictive–corrective particle-based fluid simulation on GPU

The Visual Computer, 2022

The use of particles-based simulations to produce fluid animations is nowadays a frequently used ... more The use of particles-based simulations to produce fluid animations is nowadays a frequently used method by both the industrial and research sectors. Although there are many variations of the smoothed particle hydrodynamics (SPH) algorithm currently being used, they all have the common characteristic of being highly parallel in nature. They are therefore frequently implemented on graphics processing units (GPUs) to benefit of high computation capacities of modern GPUs. However, such optimizations require specific optimizations to make use of the full capacity of the GPU, with sometimes optimizations being contradictory to optimizations used in CPU implementations. In this paper, we explored various optimizations on a GPU implementation of a recent particle-based fluid simulation algorithm using an iterative pressure solver. In particular, we focused on CPU optimizations that have not been thoroughly studied for GPU implementations: the indexing for the neighbor’s structure, the frequency of the sorting of the fluid particles, the use of lookup tables for the kernel function computations and the use of a warm-start to improve the performance of the iterative pressure solver. We show that some of these optimizations are only effective for very specific hardware configurations and sometimes even impact the performance negatively. We also show that the warm-start reduces the computation time but introduces a cyclic instability in the simulation. We propose a solution to reduce this instability without requiring to modify the implementation of the fluid algorithm.

2020 International Joint Conference on Neural Networks (IJCNN), 2020

Automatic Hand detection in vehicles achieved best performances using deep learning approach. MS-... more Automatic Hand detection in vehicles achieved best performances using deep learning approach. MS-FRCNN and MS-RFCN are the models that achieve the state of the art results on the viva hand detection challenge. These models, unfortunately, require important resources to operate (more than 7 Giga bytes GPU) which presents an important limit for their large scale applications. In this paper, we address the viva challenge of hand detection in vehicles using simpler models and low computer components (2 Giga bytes GPU). We found that the ZF model is highly adequate for the problem and more importantly, it overpasses remarkably the precisions of complex model of literature MS- FRCNN. The experiment result of ZF model on the viva database shows (AP = 91.5%, AR = 85.7% at level one and AP = 80.5%, AR = 69.5% and Fps = 0.4 s at level two in measures of Average Precision (AP), Average Recall(AR) and Frame per seconds (Fps). These results of ZF reached the fourth place on the viva hand detecti...

Nous presentons une approche pour la reconnaissance d'expressions corporelles qui accompagnen... more Nous presentons une approche pour la reconnaissance d'expressions corporelles qui accompagnent le mouvement d'un personnage reel. L'idee developpee dans cet article s'inspire de travaux en syntheses d'animations. Notre sys-teme se base sur l'analyse de la difference obtenue dans le domaine frequentiel entre un mouvement expressif et un mouvement neutre. Dans ce papier, nous introduisons la notion de mouvement neutre et proposons un algorithme qui permet de l'obtenir par la minimisation d'une fonction de cout global. Le cout global tient compte de la distance parcourue par les differentes articulations et l'accelera-tion de chaque articulation au cours du mouvement. Nous avons evalue notre methode sur differentes bases de don-nees contenant des mouvements et expressions corporelles heterogenes. Les resultats obtenus sont encourageants et confirment la validite de notre demarche.