6.1.1 AsymGait

• Name:
  Asymmetry index for clinical gait analysis based on depth images
• Keywords:
  Motion analysis, Kinect, Clinical analysis
• Scientific Description:
  The system uses depth images delivered by the Microsoft Kinect to retrieve the gait cycles first. To this end it is based on a analyzing the knees trajectories instead of the feet to obtain more robust gait event detection. Based on these cycles, the system computes a mean gait cycle model to decrease the effect of noise of the system. Asymmetry is then computed at each frame of the gait cycle as the spatial difference between the left and right parts of the body. This information is computed for each frame of the cycle.
• Functional Description:
  AsymGait is a software package that works with Microsoft Kinect data, especially depth images, in order to carry-out clinical gait analysis. First it identifies the main gait events using the depth information (footstrike, toe-off) to isolate gait cycles. Then it computes a continuous asymmetry index within the gait cycle. Asymmetry is viewed as a spatial difference between the two sides of the body.
• Contact:
  Franck Multon
• Participants:
  Edouard Auvinet, Franck Multon

6.1.2 Cinematic Viewpoint Generator

• Keyword:
  3D animation
• Functional Description:
  The software, developed as an API, provides a mean to automatically compute a collection of viewpoints over one or two specified geometric entities, in a given 3D scene, at a given time. These viewpoints satisfy classical cinematographic framing conventions and guidelines including different shot scales (from extreme long shot to extreme close-up), different shot angles (internal, external, parallel, apex), and different screen compositions (thirds,fifths, symmetric of di-symmetric). The viewpoints allow to cover the range of possible framings for the specified entities. The computation of such viewpoints relies on a database of framings that are dynamically adapted to the 3D scene by using a manifold parametric representation and guarantee the visibility of the specified entities. The set of viewpoints is also automatically annotated with cinematographic tags such as shot scales, angles, compositions, relative placement of entities, line of interest.
• Contact:
  Marc Christie
• Participants:
  Christophe Lino, Emmanuel Badier, Marc Christie
• Partners:
  Université d'Udine, Université de Nantes

6.1.3 CusToM

• Name:
  Customizable Toolbox for Musculoskeletal simulation
• Keywords:
  Biomechanics, Dynamic Analysis, Kinematics, Simulation, Mechanical multi-body systems
• Scientific Description:

  The present toolbox aims at performing a motion analysis thanks to an inverse dynamics method.

  Before performing motion analysis steps, a musculoskeletal model is generated. Its consists of, first, generating the desire anthropometric model thanks to models libraries. The generated model is then kinematical calibrated by using data of a motion capture. The inverse kinematics step, the inverse dynamics step and the muscle forces estimation step are then successively performed from motion capture and external forces data. Two folders and one script are available on the toolbox root. The Main script collects all the different functions of the motion analysis pipeline. The Functions folder contains all functions used in the toolbox. It is necessary to add this folder and all the subfolders to the Matlab path. The Problems folder is used to contain the different study. The user has to create one subfolder for each new study. Once a new musculoskeletal model is used, a new study is necessary. Different files will be automaticaly generated and saved in this folder. All files located on its root are related to the model and are valuable whatever the motion considered. A new folder will be added for each new motion capture. All files located on a folder are only related to this considered motion.

• Functional Description:
  Inverse kinematics Inverse dynamics Muscle forces estimation External forces prediction
• Publications:
  hal-02268958, hal-02088913, hal-02109407, hal-01904443, hal-02142288, hal-01988715, hal-01710990
• Contact:
  Charles Pontonnier
• Participants:
  Antoine Muller, Charles Pontonnier, Georges Dumont, Pierre Puchaud, Anthony Sorel, Claire Livet, Louise Demestre

6.1.4 Directors Lens Motion Builder

• Keywords:
  Previzualisation, Virtual camera, 3D animation
• Functional Description:
  Directors Lens Motion Builder is a software plugin for Autodesk's Motion Builder animation tool. This plugin features a novel workflow to rapidly prototype cinematographic sequences in a 3D scene, and is dedicated to the 3D animation and movie previsualization industries. The workflow integrates the automated computation of viewpoints (using the Cinematic Viewpoint Generator) to interactively explore different framings of the scene, proposes means to interactively control framings in the image space, and proposes a technique to automatically retarget a camera trajectory from one scene to another while enforcing visual properties. The tool also proposes to edit the cinematographic sequence and export the animation. The software can be linked to different virtual camera systems available on the market.
• Contact:
  Marc Christie
• Participants:
  Christophe Lino, Emmanuel Badier, Marc Christie
• Partner:
  Université de Rennes 1

6.1.5 Kimea

• Name:
  Kinect IMprovement for Egronomics Assessment
• Keywords:
  Biomechanics, Motion analysis, Kinect
• Scientific Description:
  Kimea consists in correcting skeleton data delivered by a Microsoft Kinect in an ergonomics purpose. Kimea is able to manage most of the occlultations that can occur in real working situation, on workstations. To this end, Kimea relies on a database of examples/poses organized as a graph, in order to replace unreliable body segments reconstruction by poses that have already been measured on real subject. The potential pose candidates are used in an optimization framework.
• Functional Description:
  Kimea gets Kinect data as input data (skeleton data) and correct most of measurement errors to carry-out ergonomic assessment at workstation.
• Publications:
  hal-01612939v1, hal-01393066v1, hal-01332716v1, hal-01332711v2, hal-01095084v1
• Contact:
  Franck Multon
• Participants:
  Franck Multon, Hubert Shum, Pierre Plantard
• Partner:
  Faurecia

6.1.6 Populate

• Keywords:
  Behavior modeling, Agent, Scheduling
• Scientific Description:

  The software provides the following functionalities:

  - A high level XML dialect that is dedicated to the description of agents activities in terms of tasks and sub activities that can be combined with different kind of operators: sequential, without order, interlaced. This dialect also enables the description of time and location constraints associated to tasks.

  - An XML dialect that enables the description of agent's personal characteristics.

  - An informed graph describes the topology of the environment as well as the locations where tasks can be performed. A bridge between TopoPlan and Populate has also been designed. It provides an automatic analysis of an informed 3D environment that is used to generate an informed graph compatible with Populate.

  - The generation of a valid task schedule based on the previously mentioned descriptions.

  With a good configuration of agents characteristics (based on statistics), we demonstrated that tasks schedules produced by Populate are representative of human ones. In conjunction with TopoPlan, it has been used to populate a district of Paris as well as imaginary cities with several thousands of pedestrians navigating in real time.

• Functional Description:
  Populate is a toolkit dedicated to task scheduling under time and space constraints in the field of behavioral animation. It is currently used to populate virtual cities with pedestrian performing different kind of activities implying travels between different locations. However the generic aspect of the algorithm and underlying representations enable its use in a wide range of applications that need to link activity, time and space. The main scheduling algorithm relies on the following inputs: an informed environment description, an activity an agent needs to perform and individual characteristics of this agent. The algorithm produces a valid task schedule compatible with time and spatial constraints imposed by the activity description and the environment. In this task schedule, time intervals relating to travel and task fulfillment are identified and locations where tasks should be performed are automatically selected.
• Contact:
  Fabrice Lamarche
• Participants:
  Carl-Johan Jorgensen, Fabrice Lamarche

6.1.7 PyNimation

• Keywords:
  Moving bodies, 3D animation, Synthetic human
• Scientific Description:
  PyNimation is a python-based open-source (AGPL) software for editing motion capture data which was initiated because of a lack of open-source software enabling to process different types of motion capture data in a unified way, which typically forces animation pipelines to rely on several commercial software. For instance, motions are captured with a software, retargeted using another one, then edited using a third one, etc. The goal of Pynimation is therefore to bridge the gap in the animation pipeline between motion capture software and final game engines, by handling in a unified way different types of motion capture data, providing standard and novel motion editing solutions, and exporting motion capture data to be compatible with common 3D game engines (e.g., Unity, Unreal). Its goal is also simultaneously to provide support to our research efforts in this area, and it is therefore used, maintained, and extended to progressively include novel motion editing features, as well as to integrate the results of our research projects. At a short term, our goal is to further extend its capabilities and to share it more largely with the animation/research community.
• Functional Description:

  PyNimation is a framework for editing, visualizing and studying skeletal 3D animations, it was more particularly designed to process motion capture data. It stems from the wish to utilize Python’s data science capabilities and ease of use for human motion research.

  In its version 1.0, Pynimation offers the following functionalities, which aim to evolve with the development of the tool : - Import / Export of FBX, BVH, and MVNX animation file formats - Access and modification of skeletal joint transformations, as well as a certain number of functionalities to manipulate these transformations - Basic features for human motion animation (under development, but including e.g. different methods of inverse kinematics, editing filters, etc.). - Interactive visualisation in OpenGL for animations and objects, including the possibility to animate skinned meshes

• Authors:
  Ludovic Hoyet, Robin Adili, Benjamin Niay, Alberto Jovane
• Contact:
  Ludovic Hoyet

6.1.8 The Theater

• Keywords:
  3D animation, Interactive Scenarios
• Functional Description:
  The Theater is a software framework to develop interactive scenarios in virtual 3D environements. The framework provides means to author and orchestrate 3D character behaviors and simulate them in real-time. The tools provides a basis to build a range of 3D applications, from simple simulations with reactive behaviors, to complex storytelling applications including narrative mechanisms such as flashbacks.
• Contact:
  Marc Christie
• Participant:
  Marc Christie

6.2.1 Immerstar Platform

Participants: Georges Dumont [contact], Ronan Gaugne, Anthony Sorel, Richard Kulpa.

With the two platforms of virtual reality, Immersia) and Immermove Immermove, grouped under the name Immerstar, the team has access to high level scientific facilities. This equipment benefits the research teams of the center and has allowed them to extend their local, national and international collaborations. The Immerstar platform was granted by a Inria CPER funding for 2015-2019 that enabled important evolutions of the equipment. In 2016, the first technical evolutions have been decided and, in 2017, these evolutions have been implemented. On one side, for Immermove, the addition of a third face to the immersive space, and the extension of the Vicon tracking system have been realized and continued this year with 23 new cameras. And, on the second side, for Immersia, the installation of WQXGA laser projectors with augmented global resolution, of a new tracking system with higher frequency and of new computers for simulation and image generation in 2017. In 2018, a Scale One haptic device has been installed. It allows, as in the CPER proposal, one or two handed haptic feedback in the full space covered by Immersia and possibility of carrying the user. Based on these supports, in 2020, we participated to a PIA3-Equipex+ proposal. This proposal CONTINUUM involves 22 partners, has been succesfully evaluated and will be granted. The CONTINUUM project will create a collaborative research infrastructure of 30 platforms located throughout France, to advance interdisciplinary research based on interaction between computer science and the human and social sciences. Thanks to CONTINUUM, 37 research teams will develop cutting-edge research programs focusing on visualization, immersion, interaction and collaboration, as well as on human perception, cognition and behaviour in virtual/augmented reality, with potential impact on societal issues. CONTINUUM enables a paradigm shift in the way we perceive, interact, and collaborate with complex digital data and digital worlds by putting humans at the center of the data processing workflows. The project will empower scientists, engineers and industry users with a highly interconnected network of high-performance visualization and immersive platforms to observe, manipulate, understand and share digital data, real-time multi-scale simulations, and virtual or augmented experiences. All platforms will feature facilities for remote collaboration with other platforms, as well as mobile equipment that can be lent to users to facilitate onboarding. The kick-off meeting of continuum will be held in 2022, Januray the 14th.

7.2.1 Binary Graph Descriptor for Robust Relocalization on Heterogeneous Data

Participants: Marc Christie [contact], Xi Wang.

Motion analysis in open and uncontrolled environments remains a challenging task. We here address the specific issue of robust relocalization of cameras that finds many applications in SLAM system (Simultaneous Localization and Mapping): extracting camera trajectories from movies, robust augmented reality, motion matching, etc. In 2021, we proposed a novel binary graph descriptor to improve loop detection for visual SLAM systems. Our contribution is twofold: i) a graph embedding technique for generating binary descriptors which conserve both spatial and histogram information extracted from images; ii) a generic mean of combining multiple layers of heterogeneous data into the proposed binary graph descriptor, coupled with a matching and geometric checking method. We also introduce an implementation of our descriptor into an incremental Bag-of-Words (iBoW) structure that improves efficiency and scalability, and propose a method to interpret Deep Neural Network (DNN) results. We evaluate our system on synthetic and real datasets across different lighting and seasonal conditions. The proposed method outperforms state-of-the-art loop detection frameworks in terms of relocalization precision and computational performance, as well as displays high robustness against cross-condition datasets. Paper and results are reported in 30.

7.2.2 TT SLAM: Dense Monocular SLAM for Planar Environments

Participants: Marc Christie [contact], Xi Wang.

As an addition to 30, we also report our 2021 work on a novel visual SLAM method with dense planar reconstruction using a monocular camera: TT-SLAM. The method exploits planar template-based trackers (TT) to compute camera poses and reconstructs a multi planar scene representation. Multiple homographies are estimated simultaneously by clustering a set of template trackers supported by superpixelized regions. Compared to RANSAC-based multiple homographies method, data association and keyframe selection issues are handled by the continuous nature of template trackers. A non-linear optimization process is applied to all the homographies to improve the precision in pose estimation. Experiments show that the proposed method outperforms RANSAC-based multiple homographies method as well as other dense method SLAM techniques such as LSD-SLAM or DPPTAM. It competes with keypoint-based techniques like ORB-SLAM while providing dense planar reconstructions of the environment. The work is reported in 41.

7.2.3 Analyzing visual attention in movies

Participants: Marc Christie [contact], Alexandre Bruckert.

In the overarching objective of better understanding movies through their high-level visual characteristics (camera motions, camera angles, scene layouts and mise-en-scene), we have been trying to search for better automated visual saliency techniques. Indeed, the way a film director plays with the spectator gaze is a key stylistic feature. A first work consisted in better estimating this visual saliency by improving existing techniques through well-designed loss functions. Indeed, deep learning techniques are widely used to model human visual saliency, to such a point that state-of-the-art performances are now only attained by deep neural networks. However, one key part of a typical deep learning model is often neglected when it comes to modeling visual saliency: the choice of the loss function.

In this work, we explored some of the most popular loss functions that are used in deep saliency models. We demonstrate that on a fixed network architecture, modifying the loss function can significantly improve (or depreciate) the results, hence emphasizing the importance of the choice of the loss function when designing a model. We also evaluate the relevance of new loss functions for saliency prediction inspired by metrics used in style-transfer tasks. Finally, we show that a linear combination of several well-chosen loss functions leads to significant improvements in performance on different datasets as well as on a different network architecture, thus demonstrating the robustness of a combined metric.

By building on our progressive understanding of saliency in movies, we actually demonstrated that current state of the art visual saliency techniques, trained on non-cinematographic contents, fail to capture the real spectator gaze displacements in a number of situations. Indeed, in the process of making a movie, directors constantly care about where the spectator will look on the screen. Shot composition, framing, camera movements or editing are tools commonly used to direct attention. In order to provide a quantitative analysis of the relationship between those tools and gaze patterns, we propose a new eye-tracking database, containing gaze pattern information on movie sequences, as well as editing annotations. We show how state-of-the-art computational saliency techniques behave on this dataset. In this work, we expose strong links between movie editing and spectators scanpaths, and open several leads on how the knowledge of editing information could improve human visual attention modeling for cinematic content49. The dataset generated and analysed during the current study is available at https://­github.­ com/­abruckert/­eye_tracking_filmmaking.

7.2.4 High-Level Features for Movie Style Understanding

Participants: Marc Christie [contact], Robin Courant.

After identifying the lack of high-level feature extraction techniques for movies, we proposed a new set of stylistic features (character pose, camera pose, scene depth, focus map, frame layering and camera motion). Automatically analysing such stylistic features in movies is a challenging task, as it requires an in-depth knowledge of cinematography. In the literature, only a handful of methods explore stylistic feature extraction, and they typically focus on limited low-level image and shot features (colour histograms, average shot lengths or shot types, amount of camera motion). However, it only capture a subset of the stylistic features which help to characterise a movie (e.g. black and white vs. coloured, or film editing). To this end, we systematically explore seven high-level features for movie style analysis: character segmentation, pose estimation, depth maps, focus maps, frame layering, camera motion type and camera pose. Our findings show that low-level features remain insufficient for movie style analysis, while high-level features seem promising. These results are reported in 43 and received the best paper award at the ICCV workshop on AI for Creative Video Editing and Understanding.

7.2.5 Camera Keyframing with Style and Control

Participants: Marc Christie [contact], Xi Wang.

We present a novel technique that enables 3D artists to synthesize camera motions in virtual environments following a camera style, while enforcing user-designed camera keyframes as constraints along the sequence (see figure 4). To solve this constrained motion in-betweening problem, we design and train a camera motion generator from a collection of temporal cinematic features (camera and actor motions) using a conditioning on target keyframes. We further condition the generator with a style code to control how to perform the interpolation between the keyframes. Style codes are generated by training a second network that encodes different camera behaviors in a compact latent space, the camera style space. Camera behaviors are defined as temporal correlations between actor features and camera motions and can be extracted from real or synthetic film clips. We further extend the system by incorporating a fine control of camera speed and direction via a hidden state mapping technique. We evaluate our method on two aspects: i) the capacity to synthesize style-aware camera trajectories with user defined keyframes; and ii) the capacity to ensure that in-between motions still comply with the reference camera style while satisfying the keyframe constraints. As a result, our system is the first style-aware keyframe in-betweening technique for camera control that balances style-driven automation with precise and interactive control of keyframes.

7.2.6 Real-Time Cinematic Tracking of Targets in Dynamic Environments

Participants: Marc Christie [contact], Ludovic Burg.

7.2.7 FaceTuneGAN: Face Autoencoder for Convolutional Expression Transfer Using Neural Generative Adversarial Networks

Participants: Franck Multon [contact], Nicolas Olivier.

Our network architecture. The encoders extract and compress the features of their input into low-dimensional content and style vectors. A decoder reconstructs a mesh from these compact representations. The style information is passed to the decoder through AdaIN normalization layers.

In this work 52, we present FaceTuneGAN, a new 3D face model representation decomposing and encoding separately facial identity and facial expression. We propose a first adaptation of image-to-image translation networks, that have successfully been used in the 2D domain, to 3D face geometry. Leveraging recently released large face scan databases, a neural network has been trained to decouple factors of variations with a better knowledge of the face, enabling facial expressions transfer and neutralization of expressive faces. Specifically, we design an adversarial architecture adapting the base architecture of FUNIT and using SpiralNet++ for our convolutional and sampling operations. Using two publicly available datasets (FaceScape and CoMA), FaceTuneGAN has a better identity decomposition and face neutralization than state-of-the-art techniques. The architecture is depicted in Figure 5. It also outperforms classical deformation transfer approach by predicting blendshapes closer to ground-truth data and with less of undesired artifacts due to too different facial morphologies between source and target.

7.2.8 Contact Preserving Shape Transfer For Rigging-Free Motion Retargeting

Participants: Franck Multon [contact], Jean Basset, Adnane Boukhayma.

In 2018, we introduced the idea of context graph to capture the relationship between body parts surfaces and enhance the quality of the motion retargetting problem. Hence, it becomes possible to retarget the motion of a source character to a target one while preserving the topological relationship between body parts surfaces. However this approach implies to strictly satisfy distance constraints between body parts, whereas some of them could be relaxed to preserve naturalness. In 2019, we introduced a new paradigm based on transfering the shape instead of encoding the pose constraints to tackle this problem for isolated poses.

In 2020, we extended this approach to handle continuity in motion, and non-human characters. this idea resulted from the collaboration with the Morpheo Inria team in Grenoble, in the context of the IPL AVATAR project. The proposed approaches were based on a nonlinear optimization framework, which involved manually editing the constraints to satisfy, and required a huge amount of computation time.

In 32, to achieve the deformation transfer, we proposed a neural encoder-decoder architecture where only identity information is encoded and where the decoder is conditioned on the pose. We use pose independent representations, such as isometry-invariant shape characteristics, to represent identity features. The model is depicted in Figure 6. Our model uses these features to supervise the prediction of offsets from the deformed pose to the result of the transfer. We show experimentally that our method outperforms state-of-the-art methods both quantitatively and qualitatively, and better generalizes better to poses not seen during training. We also introduce a fine-tuning step that allows to obtain competitive results for extreme identities, and allows to transfer simple clothing.

Overview of the proposed approach. The encoder (green) generates an identity code for the target. We feed this code to the decoder (red) along with the source, which is concatenated with the decoder features at all resolution stages. The decoder finally outputs per vertex offsets from the input source towards the identity transfer result.

7.2.9 Perception of Motion Variations in Large-Scale Virtual Human Crowds

Participants: Robin Adili [contact], Ludovic Hoyet [contact], Benjamin Niay, Anne-Hélène Olivier, Katja Zibrek.

Virtual human crowds are regularly featured in movies and video games. With a large number of virtual characters, each behaving in their own way, spectacular scenes can be produced. The more diverse the characters and their behaviors are, the more realistic the virtual crowd is expected to be perceived. Hence, creating virtual crowds is a trade-off between the cost associated with acquiring more diverse assets, namely more virtual characters with their animations, and achieving better realism. In this work 31, conducted in collaboration with Julien Pettré from the Rainbow team, our focus is on the perceived variety in virtual crowd character motions. We present an experiment exploring whether observers are able to identify virtual crowds including motion clones in the case of large-scale crowds (from 250 to 1000 characters), see figure 7. As it is not possible to acquire individual motions for such numbers of characters, we rely on a state-of-the-art motion variation approach to synthesize unique variations of existing examples for each character in the crowd. Participants then compared pairs of videos, where each character was animated either with a unique motion or using a subset of these motions. Our results show that virtual crowds with more than two motions (one per gender) were perceptually equivalent, regardless of their size. We believe these findings can help create efficient crowd applications, and are an additional step into a broader understanding of the perception of motion variety.

7.2.10 Interaction Fields: Sketching Collective Behaviours.

Participants: Marc Christie [contact], Adèle Colas [contact], Ludovic Hoyet [contact], Anne-Hélène Olivier [contact], Katja Zibrek.

Many applications of computer graphics, such as cinema, video games, virtual reality, training scenarios, therapy, or rehabilitation, involve the design of situations where several virtual humans are engaged. In applications where a user is immersed in the virtual environment, the (collective) behavior of these virtual humans must be realistic to improve the user's sense of presence. While expressive behaviour appears to be a crucial aspect as part of the realism, collective behaviours simulated using typical crowd simulation models (e.g., social forces, potential fields, velocity selection, vision-based) usually lack expressiveness and do not allow to capture more subtle scenarios (e.g., a group of agents hiding from the user or blocking his/her way), which require the ability to simulate complex interactions. As subtle and adaptable collective behaviours are not easily modeled, there is therefore a need for more intuitive ways to design such complex scenarios. In 2020, we proposed a novel approach to sketch such interactions to define collective behaviours, which we called Interaction Fields, in collaboration with Julien Pettré and Claudio Pacchierotti in Rainbow team. Although other sketch-based approaches exist, these usually focus on goal-oriented path planning, and not on modelling social or collective behaviour. In comparison, our approach is based on a user-friendly application enabling users to draw target interactions between agents through intuitive vector fields (Figure 8). This year, we therefore extended this work to include novel features facilitating the design of expressive and collective behaviours, and conducted an experiment evaluating the usability of the approach. By considering more generic and dynamic situations, we demonstrate that we can design diversified and subtle interactions, which so far have mostly focused on predefined static scenarios.

7.2.11 Reactive Virtual Agents: A Viewpoint-Driven Approach for Bodily Nonverbal Communication

Participants: Marc Christie [contact], Ludovic Hoyet [contact], Alberto Jovane [contact], Anne-Hélène Olivier [contact], Pierre Raimbaud, Katja Zibrek.

Intelligent Virtual Agents (IVA) should be designed to perform realistic and expressive interactions that convey emotions and express personalities. These are notably expressed by humans through nonverbal behaviours in everyday interactions. Thus, the ability to perform nonverbal communication is paramount for IVA design. Nonverbal communication, notably reflected by proxemics and kinesics, contributes to the social realism of IVA interactions. In this regard, user studies have been conducted to evaluate generative models of IVAs’ expressive facial motions or body motions. These last ones in particular – divided into postures and gestures – appear to be the key to expressivity in nonverbal communication between humans, and with IVAs. Nonetheless, ensuring the expressivity of IVAs' actions is not the only challenge related to bodily nonverbal communication. IVAs should also embed the ability to react to humans or other IVAs through adapted bodily communication.

Relying on the perception-action loop involved during human interactions, this work 39 conducted in collaboration with Julien Pettré and Claudio Pacchierotti from the Rainbow team, introduces a new concept to design IVAs with reactive body motions based on a viewpoint-driven approach. We present a new paradigm for reactive behaviours simulation, based on the visual analysis of body movements, see figure 9. In this short paper, we focus on situations where the poses of all agents are known. However, the same idea can easily be extended to more complex scenarios where only partial information about the agents' poses is available. Our viewport-driven approach first analyses the observed IVA's motions from the viewpoint of the observer IVA, and then adjusts the observer IVA's reactions.

7.3.1 Biofidelity in VR

Participants: Simon Hilt [contact], Georges Dumont [contact], Charles Pontonnier [contact].

Virtual environments (VE) and haptic interfaces (HI) tend to be introduced as virtual prototyping tools to assess ergonomic features of workstations. These approaches are cost-effective and convenient since working directly on the Digital Mock-Up. It thus allows low-cost solutions for studying the ergonomics of workstations upstream of the design chain, thanks to the interaction of an operator and a digital mock-up in a VE. This is preferred solution to constructing a physical mock-up in a Real Environment (RE). However it can be usable only if the ergonomic conclusions made from the VE are similar to the ones you would make in the real world. The focus was put on the evaluation of the impact of visual and haptic renderings in terms of biomechanical fidelity for pick-and-place tasks. We developed an original setup 10, enabling to separate the effect of visual and haptic renderings in the scene. This allows us to investigate individual and combined effects of those modes of immersion and interaction on the biomechanical behavior of the subject during the task. We focused particularly on the mixed effects of the renderings on the biomechanical fidelity of the system to simulate pick-and-place tasks. Fourteen subjects performed time-constrained pick-and-place tasks in RE and VE with a real and a virtual, haptic driven object at three different speeds. Motion of the hand and muscles activation of the upper limb were recorded. A questionnaire assessed subjectively discomfort and immersion. The results revealed significant differences between measured indicators in RE and VE and with real and virtual object. Objective and subjective measures indicated higher muscle activity and higher length of the hand trajectories in VE and with HI. Another important element is that no cross effect between haptic and visual rendering was reported. Theses results confirmed that such systems should be used with caution for ergonomics evaluation, especially when investigating postural and muscle quantities as discomfort indicators. The last of this part lies in an experimental setup easily replicable to asses more systematically the biomechanical fidelity of virtual environments for ergonomics purposes 20. Haptic feedback is a way to interact with the digital Digital Mock-Up but the control of the haptic device may affect the feedback. We proposed to evaluate the biomechanical fidelity of a pick-and-place task interacting with a 6 degrees of freedom HI and a dedicated inertial and viscous friction compensation algorithm. The proposed work shows that when manipulating such low masses in the proposed setup (see Figure 11) the subjective feeling (obtained by a questionnaire) of the user using the HI does not correspond to the muscle activity. Such a result is fundamental to classify what can be transferred from virtual to real at a biomechanical level.

Top left: Experimental setup with two targets for pick-and-place tasks. A cylindrical obstacle 8 cm in diameter and 10 cm high was placed between the targets. The HI (Virtuose6D) was placed close by so that the movement was within its range of accessibility. The subject was equipped with a Vive tracker to record the position and orientation of his hand and with a myo bracelet to record the electromyographic activity of his arm. Finally, the subject was equipped with an HTC Vive Pro helmet during the virtual immersion where the virtual scene was projected. Bottom left: This virtual scene included the same elements as the real scene except for the HI with the addition of a virtual hand collocated thanks to the Vive Tracker. Right: Illustration of the four combinations of immersion/interaction modes.

7.3.2 Towards "Avatar-Friendly" 3D Manipulation Techniques: Theoretical Background and Practical Design Guidelines

Participants: Diane Dewez [contact], Ludovic Hoyet [contact].

Avatars, the users' virtual representations, are becoming ubiquitous in virtual reality applications. In this context, the avatar indeed becomes the medium which enables users to manipulate objects in the virtual environment. It also becomes the users' main spatial reference, which can not only alter their interaction with the virtual environment, but also the perception of themselves. In this work 35, conducted in collaboration with Ferran Argelaguet and Anatole Lécuyer from the Hybrid team, we review and analyse the current state-of-the-art for 3D object manipulation and the sense of embodiment. Our analysis is twofold. First, we discuss the impact that the avatar can have on object manipulation. Second, we discuss how the different components of a manipulation technique (i.e. input, control and feedback) can influence the user’s sense of embodiment. Throughout the analysis, we crystallise our discussion with practical guidelines for VR application designers and we propose several research topics towards avatar-friendly manipulation techniques.

7.3.3 Does virtual threat harm VR experience?: Impact of threat occurrence and repeatability on virtual embodiment and threat response

Participants: Rebecca Fribourg [contact], Ludovic Hoyet [contact].

This work 18, conducted in collaboration with Ferran Argelaguet and Anatole Lécuyer from the Hybrid team, extends our previous work to explore the potential impact of threat occurrence and repeatability on users' Sense of Embodiment (SoE) and threat response 12. To that aim, we conducted an experiment in which participants were embodied in a virtual avatar, and performed a task in which a threat towards the virtual body was introduced a first time, then repeated several times through the experiment (in total 5 times). The SoE of participants as well as their subjective response to the threat were assessed through subjective questionnaires before the introduction of the threat, after a first introduction of the threat and at the end of the experiment. A control group did the same experiment with no threat introduced during the task. The main findings of our experiment are that the introduction of a threat does not alter users' SoE but might change their behaviour while performing a task after the threat occurrence. In addition, threat repetitions did not show any effect on users' subjective SoE, or subjective and objective responses to threat. Taken together, our results suggest that embodiment studies should expect potential change in participants behaviour while doing a task after a threat was introduced, but that threat introduction and repetition do not seem to impact the subjective measure of the SoE (subective ratings) nor the objective measure of the SoE (physical reaction to threat).

7.3.4 May I sit next to you? The effect of motion quality of virtual agents on the proximity in virtual reality

Participants: Vincent Etien [contact], Ludovic Hoyet [contact], Katja Zibrek [contact].

Proximity is a useful measure when investigating the perception of virtual agents in virtual reality. The measure can be used to assess the level of comfort people feel with virtual agents and is a sign of social presence with the agent. In this work, conducted in collaboration with Radoslaw Sterna from Jagiellonian University (Krakow, Poland), we investigated if the motion quality of agents impacts the proximity distance towards them. We created three levels of motion quality, by manipulating different body and facial animation techniques. We also introduced a new way of measuring proximity by asking the participants, embodied in a virtual waiting room, to sit on a chair as close as possible to the virtual agent (see Figure 13. We also asked them to recognise the emotional expression of the agent, rate their uncanniness traits (appeal, eeriness, familiarity), and assess their level of motion realism. Emotion recognition and motion realism detection showed that participants could recognise emotions and the type of motion quality as predicted. Our main result is that motion quality as well as eeriness are related to proximity.

7.3.5 Exploring behaviour towards avatars and agents in immersive virtual environments with mixed-agency interactions

Participants: Ludovic Hoyet [contact], Anne-Hélène Olivier [contact], Katja Zibrek.

Immersive virtual environments (IVEs) in which multiple users navigate by walking and interact with each other in natural ways are perfectly suited for team applications from training to recreation. At the same time, they can solve scheduling conflicts by employing virtual agents in place of missing team members or additional participants of a scenario. While this idea has been long discussed in IVEs research, there are no prior publications on social interactions in systems with multiple embodied users and agents. This work 38, conducted in collaboration with Julien Pettré from the Rainbow team, presents an experiment at a work-in-progress stage that addresses the impact of perceived agency and control of a virtual character in a collaborative scenario, with two embodied users and one virtual agent. Our future study will investigate whether users treat avatars and agents differently within a mixed-agency scenario, analysing several behavioural metrics and self-report of participants.

7.3.6 Understanding, Modeling and Simulating Unintended Positional Drift during Repetitive Steering Navigation Tasks in Virtual Reality

Participants: Anne-Hélène Olivier [contact], Hugo Brument.

This work was performed in collaboration with Ferran Argelaguet and Maud Marchal from Hybrid and Rainbow teams 12.

Virtual steering techniques enable users to navigate in larger Virtual Environments (VEs) than the physical workspace available. Even though these techniques do not require physical movement of the users (e.g. using a joystick and the head orientation to steer towards a virtual direction), recent work observed that users might unintentionally move in the physical workspace while navigating, resulting in Unintended Positional Drift (UPD). This phenomenon can be a safety issue since users may unintentionally reach the physical boundaries of the workspace while using a steering technique. In this context, as a necessary first step to improve the design of navigation techniques minimizing the UPD, we aim here analyzing and modeling the UPD during a virtual navigation task. In particular, we characterize and analyze the UPD for a dataset containing the positions and orientations of eighteen users performing a virtual slalom task using virtual steering techniques (cf Figure 14). Participants wore a head-mounted display and had to follow three different sinusoidal-like trajectories (with low, medium and high curvature) using a torso-steering navigation technique. We analyzed the performed motions and proposed two UPD models: the first based on a linear regression analysis and the second based on a Gaussian Mixture Model (GMM) analysis. Then, we assessed both models through a simulation-based evaluation where we reproduced the same navigation task using virtual agents. Our results indicate the feasibility of using simulation-based evaluations to study UPD. We conclude with a discussion of potential applications of the results in order to gain a better understanding of UPD during steering and therefore improve the design of navigation techniques by compensating for UPD.

7.3.7 Studying the Influence of Translational and Rotational Motion on the Perception of Rotation Gains in Virtual Environments

Participants: Anne-Hélène Olivier [contact], Hugo Brument.

This work was performed in collaboration with Ferran Argelaguet and Maud Marchal from Hybrid and Rainbow teams 33.

Rotation gains in Virtual Reality (VR) enable the exploration of wider Virtual Environments (VEs) compared to the workspace users have in VR setups. The perception of these gains has been consequently explored through multiple experimental conditions in order to improve redirected navigation techniques. While most of the studies consider rotations, in which participants can rotate to the place they desire but without translational motion, we have no information about the potential impact of the translational and rotational motions on the perception of rotation gains. In this paper, we estimated the influence of these motions and compared the perceptual thresholds of rotations gains through a user study (n = 14), in which participants had to perform virtual rotation tasks at a constant rotation speed. Participants had to determine whether their virtual rotation speed was faster or slower than their real one. We varied the translational optical flow (static or forward motion), the rotational speed (20, 30, or 40 deg/s), and the rotational gain (from 0.5 to 1.5). The main results are that the rotation gains are less perceivable at lower rotation speeds and that translational motion makes detection more difficult at lower rotation speeds. Furthermore, we provides insights into the user’s gaze and body motions behaviour when exposed to rotation gains. These results contribute to the understanding of the perception of rotation gains in VEs and they are discussed to improve the implementation of rotation gains in redirection techniques.

7.4.1 Monocular Human Shape and Pose with Dense Mesh-borne Local Image Features

Participants: Adnane Boukhayma [contact], Franck Multon, Shubhendu Jena.

We propose to improve on graph convolution based approaches for human shape and pose estimation from monocular input, using pixel-aligned local image features. Given a single input color image, existing graph convolutional network (GCN) based techniques for human shape and pose estimation use a single convolutional neural network (CNN) generated global image feature appended to all mesh vertices equally to initialize the GCN stage, which transforms a template T-posed mesh into the target pose. In contrast, we propose for the first time the idea of using local image features per vertex. Figure 15 depicts the architecture of such a new solution. These features are sampled from the CNN image feature maps by utilizing pixel-to-mesh correspondences generated with DensePose. Our quantitative and qualitative results on standard benchmarks show that using local features improves on global ones and leads to competitive performances with respect to the state-of-the-art. This work was published at the IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021) 50.

7.4.2 Dual Mesh Convolutional Networks for Human Shape Correspondence

Participants: Adnane Boukhayma [contact].

Convolutional neural networks have been extremely successful for 2D images and are readily extended to handle 3D voxel data. Meshes are a more common 3D shape representation that quantize the shape surface instead of the ambient space as with voxels, hence giving access to surface properties such as normals or appearances. The formulation of deep neural networks on meshes is, however, more complex since they are irregular data structures where the number of neighbors varies across vertices. While graph convolutional networks have previously been proposed over mesh vertex data, we explore how these networks can be extended to the dual face-based representation of triangular meshes, where nodes represent triangular faces in place of vertices. In comparison to the primal vertex mesh, its face dual offers several advantages, including, importantly, that the dual mesh is regular in the sense that each triangular face has exactly three neighbors. Moreover, the dual mesh suggests the use of a number of input features that are naturally defined over faces, such as surface normals and face areas. We evaluate the dual approach on the shape correspondence task on the FAUST human shape dataset and other versions of it with varying mesh topology. While applying generic graph convolutions to the dual mesh shows already improvements over primal mesh inputs, our experiments demonstrate that building additionally convolutional models that explicitly leverage the neighborhood size regularity of dual meshes enables learning shape representations that perform on par or better than previous approaches in terms of correspondence accuracy and mean geodesic error, while being more robust to topological changes in the meshes between training and testing shapes. This work was published at the IEEE International Conference on 3D Vision (3DV 2021) and accepted as an oral presentation  40. It was a collaboration with Nivika Verma and Edmond boyer from the Morpheo team in Inria Grenoble, and Jakob Verbeek from Facebook Research.

7.4.3 Muscle path modeling for simplified musculoskeletal analysis

Participants: Claire Livet [contact], Georges Dumont [contact], Charles Pontonnier [contact].

The current study 21 aims at proposing an automatic method to design and adjust simplified muscle paths of a musculoskeletal model. These muscle paths are composed of straight-lines described by a limited set of fixed active via points and an optimization routine is developed to place these via points on the model in order to fit moment arms and musculotendon lengths input data. The method has been applied to a forearm musculoskeletal model extracted from the literature, using theoretical input data as an example. Results showed that for 75% of the muscle set, the relative root mean square error between literature theoretical data and the results from optimized muscle path was under 29.23% for moment arms and of 1.09% for musculotendon lengths. These results confirm the ability of the method to automatically generate computationally efficient muscle paths for musculoskeletal simulations. Using only via points lowers computational expense compared to paths exhibiting wrapping objects. A proper balance between computational time and anatomical realism should be found to help those models being interpreted by practitioners.

Muscle types taken into account in the muscle path modeling method developed in

7.4.4 Musculosekeletal analysis library developments

Participants: Claire Livet [contact], Georges Dumont [contact], Charles Pontonnier [contact].

Our team develops since a few years a musculoskeletal analysis toolbox (CusToM, that serves at a basis for many different PhD works now. We continue to enhance the capabilities of this toolbox by integrating PhD works. This year, we worked specifically on the development of closed loop models such as the kinematics of the shoulder complex 22 and the musculoskeletal modeling of the forearm 46. We also presented CusToM at the national conference Sciences 2024 to promote its usage in sport analysis 42.

7.4.5 pressure insoles assessment for external forces prediction

Participants: Pauline Morin [contact], Georges Dumont [contact], Charles Pontonnier [contact].

7.5.1 Early Motion Recognition

Participants: Richard Kulpa [contact], William Mocaer.

In order to produce sports training tools in virtual reality integrating strong interactions between the real and virtual athlete, it is necessary to have the earliest possible recognition of the gesture of the immersed athlete. This allows to react as soon as possible, well before the end of the action. To begin with, we propose OLT-C3D (Online Long-Term Convolutional 3D), a new architecture based on a 3D Convolutional Neural Network (3D CNN) inspired by recent spatio-temporal Convolutional Neural Networks in computer vision field, to address the complex task of early recognition of 2D handwritten gestures in real time (47). . The input signal of the gesture is converted to an image sequence along time with the trajectory history. The image sequence is passed to our 3D CNN OLT-C3D which gives a prediction at each new frame. OLT-C3D is coupled with an integrated temporal reject system to postpone the decision in time if more information is needed. Moreover, our system is end-to-end trainable, OLT-C3D and the temporal reject system are jointly trained to optimize the earliness of the decision. Our approach achieves superior performances on two complementary and freely available datasets: ILGDB and MTGSetB.

7.5.2 Diving board characterization for diving analysis

Participants: Georges Dumont [contact], Charles Pontonnier, Guillaume Nicolas, Nicolas Bideau, Louise Demestre, Pauline Morin.

7.6.1 Exoskeleton biomechanical impact assessment

Participants: Charles Pontonnier [contact], Divyaksh Subhash Chander, Simon Kirchhofer.

7.6.2 Posture Assessment and Subjective Scale Agreement in Picking Tasks with Low Masses

Participants: Franck Multon [contact], Charles Pontonnier, Georges Dumont, Olfa Haj Mahmoud.

This work aims at analyzing the relationship between postural assessment and perceived discomfort for picking tasks with low mass (up to 3kg), involving a wide range of positions/postures. In 2020, we analyzed the agreement of the different postural scores (mean value, integral value, root mean square value, weighted average time at each RULA level and the % of time per RULA level) with the subjective assessments. The statistical analysis showed no correlation between subjective and postural scores.

Based on this result, in 2021, we extended this approach by considering more variables as input (not limited to posture assessment) and replacing linear statistical relationship by non linear machine learning methods 19.

Hence, a neural networks approach has been proposed to handle various inputs such as postural, anthropometric and environmental variables in order to estimate self-reported discomfort in picking tasks. An input reduction method has been proposed, reducing the input variables to the minimum data required to estimate self-reported discomfort with similar accuracy as the neural network fed with all variables. To this end, eleven subjects carried-out picking tasks with various masses (0, 1, 3 kg) and imposed duration (5, 10 or 15 seconds). Continuous REBA score, anthropometric and environmental data were computed, and subjects’ discomfort were collected. The data set of this work consisted in the computed continuous REBA score, anthropometric, environmental data and collected subjects’ discomfort. The results showed that the correlation between the estimated and experimental tested data was equal to 0.775 when using all the 14 available variables. After data reduction, only 6 variables were left, with a very close performance when predicting discomfort. This method has the potential to support ergonomists in workstation designing processes, by adding discomfort prediction to virtual manikins’ behaviors in simulation tools.

7.7.1 Modeling physical interactions in human dense crowds: study of individual response to controlled external pushes

Participants: Charles Pontonnier [contact], Ludovic Hoyet, Anne-Hélène Olivier, Thomas Chatagnon.

This work is performed in collaboration with Julien Pettré from the Rainbow team, in the frame of the European project CrowdDNA and the PhD of Thomas Chatagnon.

Dense crowds are complex environments in which obtaining meaningful metrics to understand and predict general behavior can be really challenging. Modelling such phenomena requires to consider local interactions such as local contacts and resulting motion. Dense crowds are generally found in cultural, social or religious events (concerts, pilgrimages, giant sales...) in which many stimuli (pushes) may come from many directions on individuals, resulting in large motions of the crowd that can have tragic outcomes. At the same time, individuals may be focused on many other cognitive distractions (e.g. music, video) that may affect their response to these stimuli.

Following works on push recovery and on reaction to external perturbations, we designed this year a pilot study (Figure 18) aiming at relating the response of a subject to external perturbations for several situations, including potential distractions (by controlling the availability of some of their sensory inputs) 13.

7.7.2 Visuomotor coordinations and person-specific factors during a collision avoidance task

Participants: Anne-Hélène Olivier [contact], Armel Crétual.

This research is performed in close collaboration with Wilfrid Laurier University (Canada) in the frame of Inria Bear Associate team. We have conducted experiments to understand the effect of person-specific factors on visuomotor coordination during a collision avoidance locomotor task. Collision avoidance between two walkers requires a mutual adaptation based on visual information in order to be successful.

We first focused on age-related changes 28. Indeed these changes to visuomotor processing, kinesthetic input, and intersegmental dynamics increases the risk of collision and falls in older adults. However, few studies examine behavioural strategies in older adults during collision avoidance tasks with another pedestrian. Is there a difference between older adults’ and young adults’ collision avoidance behaviours with another pedestrian? Seventeen older adults and seventeen young adults walked at a comfortable walking speed along a 12.6 m pathway while avoiding another walker. Trials were randomized equally to include 20 interactions with the same age group and 21 interactions with the opposite age group. Minimum predicted distance (mpd) was used to characterize collision avoidance behaviours between older adults and young adults. Our results showed that older adults had riskier avoidance behaviours, passing closer to the other pedestrian compared to when two young adults were on a collision course. Whenever an older adult was on a collision course with a young adult, the young adult contributed more to the avoidance regardless of passing order. The results from the current study highlight age-related effects during a collision avoidance task in older adults resulting in risky behaviour and a potential collision. Future studies should further investigate age-related visuomotor deficits during collision avoidance tasks in cluttered environments using virtual reality in order to tease out factors that contribute most to avoidance behaviours in older adults.

Second, we considered the effect of a concussion on these visuomotor coordinations 29. Individuals who have sustained a concussion often display associated balance control deficits and visuomotor impairments despite being cleared by a physician to return to sport. Such visuomotor impairments can be highlighted in collision avoidance tasks that involves a mutual adaptation between two walkers. However, studies have yet to challenge athletes with a previous concussion during an everyday collision avoidance task, following return to sport. We then wonder whether athletes with a previous concussion display associated behavioural changes during a 90 degree collision avoidance task with an approaching pedestrian. To this end, we conducted an experiment involving thirteen athletes (ATH; 9 females) and 13 athletes with a previous concussion (CONC; 9 females, concussion <6 months). They were asked to walke at a comfortable walking speed along a 12.6 m pathway while avoiding another athlete on a 90degree collision course. Each participant randomly interacted with individuals from the same group 20 times and interacted with individuals from the opposite group 21 times. Minimum predicted distance (mpd) was used to examine collision avoidance behaviours between ATH and CONC groups. Our results showed that the overall progression of mpd(t) did not differ between groups. During the collision avoidance task, previously concussed athletes contributed less when passing second compared to their peers. When two previously concussed athletes were on a collision course, there was a greater amount of variability resulting in inappropriate adaptive behaviours. Although successful at avoiding a collision with an approaching athlete, previously concussed athletes exhibit behavioural changes manifesting in riskier behaviours. The current findings suggest that previously concussed athletes possess behavioural changes even after being cleared to returned to sport, which may increase their risk of a subsequent injury when playing.

7.7.3 Interactions between walkers in VR

Participants: Anne-Hélène Olivier [contact], Armel Crétual, Ludovic Hoyet, Benjamin Niay.

A first axis of this work concerns the effect of walkers’ path on avoidance behavior 23. Navigating crowded community spaces requires interactions with pedestrians that follow rectilinear and curvilinear trajectories. In the case of rectilinear trajectories, it has been shown that the perceived action opportunities of the walkers might be afforded based on a future distance of closest approach. However, little is known about collision avoidance behaviours when avoiding walkers that follow curvilinear trajectories. Twenty-two participants were immersed in a virtual environment and avoided a virtual human (VH) that followed either a rectilinear path or a curvilinear path with a 5 m or 10 m radius curve at various distances of closest approach. Compared to a rectilinear path (control condition), the curvilinear path with a 5 m radius yielded more collisions when the VH approached from behind the participant and more inversions when the VH approached from in-front. During each trial, the evolution of the future distance of closest approach showed similarities between rectilinear paths and curvilinear paths with a 10 m radius curve. Overall, with few collisions and few inversions of crossing order, we can conclude that participants were capable of predicting future distance of closest approach of virtual walkers that followed curvilinear trajectories. The task was solved with similar avoidance adaptations to those observed for rectilinear interactions. These findings should inform future endeavors to further understand collision avoidance strategies and the role of—for example—non-constant velocities.

In the continuity of this work, a second axis has focused on the influence of gait patterns 11. Indeed, gait patterns provide a rich source of person-specific information such as age, sex, identity, and vulnerability. However, it is unknown to what extent person-specific gait information can affect collision avoidance behaviours with an approaching “person”. We sought to determine whether young adults’ spatiotemporal avoidance behaviours were affected by changes to a virtual agent’s gait parameters (i.e., speed or trunk sway). In a virtual environment (FOVE head-mounted display; 70Hz), young adults (n=21) walked along an 18m pathway towards a goal while avoiding an approaching virtual agent. The agent’s walking speed and trunk sway magnitude were a multiples of each participant’s average speed or sway: fast (x1.5m/s), normal (matched), or slow (x0.8m/s); large (2x), normal (matched), or small (0x) respectively. The agent was non-reactive and walked straight forward at a constant speed. Participants’ kinematics were recorded (Qualisys; 120Hz) to examine avoidance behaviours of initiation of path deviation and medial-lateral clearance at the time of crossing. Statistical analysis revealed that participants initiated a path deviation (i.e., estimate of time-to-contact, TTC) significantly earlier when the agent was walking fast (M=3.93s, SD=.56) as opposed to normal (M=4.31s, SD=.33) and slow (M=4.41s, SD=.44) walking speeds. However, the agent’s trunk sway magnitudes did not affect participants’ initiation of path deviation or medial-lateral clearance at crossing. Participants appear not to use temporal information to initiate an avoidance, but rather a point in space (i.e., TTC was affected by approach speed) due to awareness that the agent was non-reactive and always approaching. The agent’s sway magnitude did not affect medial-lateral avoidance behaviours most likely because there was little observable difference between. Conceivably, the study’s environmental conditions may underrepresent people’s behaviours in real-world. Future work is needed to understand the perception of an approaching person’s gait characteristics on collision avoidance.

Illustration of the experiment in the simulator room to study the interaction between a pedestrian and self-driving and conventional cars.

Finally, in a third axis, we consider the interaction between pedestrians and autonomous vehicles 16. Self-driving vehicles are gradually becoming a reality. But the consequences of introducing such automated vehicles (AVs) into current road traffic cannot be clearly foreseen yet, especially for pedestrian safety. The present study used virtual reality to examine the pedestrians' crossing behavior in front of AVs as compared to conventional cars (CVs) (Figure 19). Thirty young (ages 21-39) and 30 older (ages 68-81) adults participated in a simulated street-crossing experiment allowing for a real walk across an experimental two-way street. Participants had to cross (or not cross) in mixed traffic conditions where highly perceptible AVs always stopped to let them cross, while CVs did not brake to give them the right of way. Available time gap (from 1 to 5 s), approach speed (30 or 50 km/h), and the lane in which the cars were approaching (near and/or far lane of the two-way street) were varied. The results revealed a significantly higher propensity to cross the street, at shorter gaps, when AVs gave way to participants in the near lane while CVs were approaching in the far lane, leading to more collisions in this condition than in the others. These risky decisions were observed for both young and older participants, but much more so for the older ones. The results also indicated hesitation to cross in front of an AV in both lanes of the two-way street, with later initiations and longer crossing times, especially for the young participants and when the AVs were approaching at a short distance and braked suddenly. This study highlights the potential risks for pedestrians of introducing AVs into current road traffic, complicating the street-crossing task for young and older people alike. Future studies should look further into the role of repeated practice and trust in AVs.

Cifre Faurecia - Monitoring of gestual efficiency at work

Participant: Franck Multon [contact], Georges Dumont, Charles Pontonnier, Olfa Haj Mahmoud.

This Cifre contract has started in September 2018 for three years and is funding the PhD thesis of Olfa Haj Mamhoud. It consists in designing new methods based on depth cameras to monitor the activity of workers in production lines, computes the potential risk of musculoskeletal disorders, and efficiency compared to reference workers. It raises several fundamental questions, such as adapting previous methods to assess the risk of musculoskeletal disorders, as they generally rely on static poses whereas the worker is performing motion. Based on previous works in the team (previous Cifre PhD thesis of Pierre Plantard) we will provide 30Hz motion capture of the worker, that will enable us to evaluate various time-dependent assessment methods.

We will also explore how to estimate joint forces based and torques on such noisy and low-sampling motion data. We will then define a new assessment method based on these forces and torques.

The Cifre contracts funds the PhD salary and 10K€ per year for the supervision and management of the PhD thesis.

Cifre InterDigitial - Adaptive Avatar Customization for Immersive Experiences

Participant: Franck Multon [contact], Ludovic Hoyet, Nicolas Olivier.

This Cifre contract has started in February 2019 for three years and is funding the PhD thesis of Nicolas Olivier. The aim of the project is to design stylized avatars of users in immersive environment and digital arts such as videogames or cinema.

To this end, we will design a pipeline from motion and shape capture of the user to the simulation of the 3D real-time and stylized avatar. It will take hairs, eyes, face, body shape and motion into account. The key idea is to stylized both appaearance and motion to make avatar better correspond to the style of the movie of immersive experience. We will carry-out perceptual studies to better understand the expectation of the users when controlling stylized avatars, to maximize embodiment. The Cifre contracts funds the PhD salary and 15K€ per year for the supervision and management of the PhD thesis. This contract is also in collaboration with Hybrid team.

Cifre InterDigitial - Learning-Based Human Character Animation Synthesis for Content Production

Participant: Ludovic Hoyet [contact], Lucas Mourot.

The overall objective of the PhD thesis of Lucas Mourot, which started in June 2020, is to adapt and improve the state-of-art on video animation and human motion modelling to develop a semi-automated framework for human animation synthesis that brings real benefits to artists in the movie and advertising industry. In particular, one objective is to leverage the use of novel learning-based approaches, in order to propose skeleton-based animation representations, as well as editing tools, in order to improve the resolution and accuracy of the produced animations, so that automatically synthetized animations might become usable in an interactive way by animation artists.

Collaboration with company SolidAnim (Bordeaux, France)

Participant: Marc Christie [contact], Xi Wang.

This contract started in November 2019 for three years. Its purpose is to explore novel means of performing depth detection for augmented reality applied to the film and broadcast industries. The grant serves to fund the PhD of Xi Wang. (160kE)

Collaboration Unreal company (Unreal Megagrant)

Participant: Marc Christie [contact].

This contract started in September 2020 for two years. The objective is to explore means of designing novel VR manipulators for character animation tasks in Unreal Engine. (70kE)

9.1.1 Inria associate team not involved in an IIL or an international program

9.2.1 FP7 & H2020 projects

9.3.1 ANR

-

9.3.2 Défi Avatar

Participants: Jean Basset [contact], Diane Dewez [contact], Rebecca Fribourg [contact], Ludovic Hoyet [contact], Franck Multon.

This project aims at design avatars (i.e., the user’s representation in virtual environments) that are better embodied, more interactive and more social, through improving all the pipeline related to avatars, from acquisition and simulation, to designing novel interaction paradigms and multi-sensory feedback. It involves 6 Inria teams (GraphDeco, Hybrid, Loki, MimeTIC, Morpheo, Potioc), Prof. Mel Slater (Uni. Barcelona), and 2 industrial partners (InterDigitak and Faurecia).

Website

9.4.1 KIMEA Cloud Project

Participants: Franck Multon [contact], Adnane Boukhayma, Shubhendu Jena.

The Region Bretagne with “Image & Réseaux” national clusters have funded the project “Kimea Cloud” (Sept 2020-June 2022). This project leaded by Moovency (start-up company created in 2018 based on a PhD from MimeTIC) aims at proposing a new system to assess painful movement at work using ligth devices, such as a simple camera. Quortex (www.quortex.io/) company provides its expertise in delivering video content on Internet to make the assessment be remote and efficient for the future customers. MimeTIC provides a robust 3D posture tracking system based on the video stream.

9.4.2 EXOSCARNE Project

Participants: Charles Pontonnier [contact], Simon Kirchhofer.

This project (2020-2023), funded by the Brittany region and endorsed by both competitive pools Valorial and EMC2, aim at designing, prototyping and commercialing a wrist exoskeleton able to help industrial butchers in their cutting and deboning tasks. It is a partnership between a R&D company called Lab4i, the mimetic team and the industrial butchery cooperl. Our role in the consortium is the development of a virtual prototyping tool based on musculoskeletal modeling to simulate the action of the exoskeleton on the wrist, and to characterize the impact of the real prototype on the action of the user thanks to full scale experimentations involving motion, force and muscle activity sensing. The project funding is about 130kE for the team

9.4.3 ADT PyToM

Participants: Charles Pontonnier [contact], Laurent Guillo, Georges Dumont, Salomé Ribault.

This project (2021-2023), funded by inria, aims at developing a Python version of our musculoskeletal library called CusToM and currently developed in Matlab. The project is also developing additional motion data entries (vision, depth cameras) in the library to enhance the usability of the analysis tools.

10.1.1 Scientific events: organisation

10.1.2 Scientific events: selection

10.1.3 Journal

10.1.4 Invited talks

• Marc Christie : Keynote talk : "Towards Computational Cinematography: what's left and what right?", ICCV Workshop on AI for Creative Video Editing and Understanding, CVEU (virtual)
• Charles Pontonnier : "La simulation musculo-squelettique pour le prototypage des exosquelettes", Colloque le soldat augmenté: Les nouvelles technologies au service de l’augmentation des performances du combattant, organisé par le CREC saint-cyr, Paris
• Anne-Hélène Olivier: "Interactions between pedestrians: from real to virtual studies", EuroXR 2021 (virtual)
• Anne-Hélène Olivier: "An interdisciplinary approach between Human Movement Sciences and Digital Sciences to understand collision avoidance behaviour between pedestrians", Rehab Week 2021 (virtual).
• Anne-Hélène Olivier: "Coordinations visuo-motrices durant la locomotion". Journées GDR IGRV (virtual).
• Franck Multon: "Les capteurs dans le sport problématiques et solutions", invité aux Journées capteurs du GDR CNRS Sport, Mars 2021 (virtual).

10.1.5 Scientific expertise

• Charles Pontonnier : expert pour l'ANR 2021, expert pour l'UNIGE
• Richard Kulpa : expert pour MITACS Canada
• Marc Christie : membre du Comité d'Evaluation 38 de l'ANR

10.1.6 Research administration

• Ludovic Hoyet is the coordinator of the Inria Research Challenge Avatar
• Franck Multon is responsible for the coordination of national Inria actions in Sports
• Franck Multon is the scientific representative of Inria in Sciences2024 group and scientific Committee
• Franck Multon is the scientific representative of Inria in the EUR Digisport steering committe and scientific committee
• Franck Multon is member of the Brittany commission of deontology
• Armel Crétual is the elected head of the Sports Sciences departement (STAPS) in University Rennes2
• Benoit Bideau is the head of the M2S Laboratory
• Benoit Bideau is vice-president of University Rennes2, in charge of the valorisation
• Benoit Bideau is the leader of the EUR DIGISPORT project
• Charles Pontonnier is the deputy director of the Mechatronics teaching and research department of ENS Rennes
• Charles Pontonnier is member of the EUR digisport pedagogical comitee
• Anne-Hélène Olivier is a co-director of the master program in adapted physical activity at the University of Rennes 2
• Richard Kulpa is the co-leader of the EUR DIGISPORT project
• Richard Kulpa is the scientific head of the EUR DIGISPORT project
• Marc Christie is the Project Coordinator of the H2020 ICT-55 project INVICTUS

10.2.1 Supervision

• PhD in progress (beginning Oct 2018, Defended December 2021) Hugo Brument, Toward user-adapted interactions techniques based on human locomotion laws for navigating in virtual environments, Anne-Hélène Olivier, Ferran Argelaguet (Hybrid team), Maud Marchal (Rainbow team).
• PhD (beginning Oct. 2018, defended Nov. 2021): Diane Dewez, Avatar-Based Interaction in Virtual Reality, Ferran Argelaguet (Hybrid team), Ludovic Hoyet, Anatole Lécuyer (Hybrid team).
• PhD (beginning September 2018, defended December 2021): Olfa Haj Mamhoud, Monitoring de l’efficience gestuelle d’opérateurs sur postes de travail, University Rennes 2, Franck Multon, Georges Dumont, Charles Pontonnier
• PhD in progress (beginning January 2019): Nils Hareng, simulation of plausible bipedal locomotion of human and non-human primate, University Rennes2, Franck Multon, & Bruno Watier (CNRS LAAS in Toulouse)
• PhD in progress (beginning January 2019): Nicolas Olivier, Adaptive Avatar Customization for Immersive Experience, Cifre InterDigital, Franck Multon, Ferran Arguelaget (Hybrid team), Quentin Avril (InterDigital), Fabien Danieau (InterDigital)
• PhD in progress (beginning September 2017): Lyse Leclerc, Intérêts dans les activités physiques du rétablissement de la fonction inertielle des membres supérieurs en cas d’amputation ou d’atrophie, Armel Crétual, Diane Haering
• PhD in progress (beginning September 2018) : Jean Basset, Learning Morphologically Plausible Pose Transfer, Inria, Edmond Boyer (Morpheo Inria Grenoble), Franck Multon
• PhD in progress (beginning December 2020): Mohamed Younes, Learning and simulating strategies in sports for VR training, University Rennes1, Franck Multon, Richard Kulpa, Ewa Kijak, Simon Malinowski
• PhD in progress (beginning September 2019) : Claire Livet, Dynamique contrainte pour l'analyse musculo-squelettique en temps rapide : vers des méthodes alternatives d'estimation des forces musculaires mises en jeu dans le mouvement humain, Ecole normale supérieure, Georges Dumont, Charles Pontonnier
• PhD in progress (beginning september 2019) : Louise Demestre, simulation MUsculo-squelettique et Structure Elastique pour le Sport (MUSES), Ecole normale supérieure, Georges Dumont, Charles Pontonnier, Nicolas Bideau, Guillaume Nicolas
• PhD in progress (beginning Oct 2018): Benjamin Niay, A framework for synthezing personalised human motions from motion capture data and perceptual information, Ludovic Hoyet, Anne-Hélène Olivier, Julien Pettré (Rainbow team).
• PhD in progress (beginning Sept 2019): Alberto Jovane, Modélisation de mouvements réactifs et comportements non verbaux pour la création d’acteurs digitaux pour la réalité virtuelle, Marc Christie, Ludovic Hoyet, Claudio Pacchierotti (Rainbow team), Julien Pettré (Rainbow team).
• PhD in progress (beginning Nov 2019): Adèle Colas, Modélisation de comportements collectifs réactifs et expressifs pour la réalité virtuelle, Ludovic Hoyet, Anne-Hélène Olivier, Claudio Pacchierotti (Rainbow team), Julien Pettré (Rainbow team).
• PhD in progress (beginning Sep 2018): Carole Puil, Impact d’une stimulation plantaire orthétique sur la posture d’individus sains et posturalement déficients au cours de la station debout, et lors de la marche, Armel Crétual, Anne-Hélène Olivier
• PhD in progress: Annabelle Limballe, Anticipation dans les sports de combat : la réalité virtuelle comme solution innovante d’entraînement, Sep. 2019, Richard Kulpa & Simon Bennett & Benoit Bideau
• PhD in progress: Alexandre Vu, Evaluation de l’influence des feedbacks sur la capacité d’apprentissage dans le cadre d’interactions complexes entre joueurs et influence de ces feedbacks en fonction de l’activité sportive , Sep. 2019, Richard Kulpa & Benoit Bideau & Anthony Sorel
• PhD in progress: William Mocaer, Réseaux de Neurones à Convolution Spatio-Temporelle pour l'analyse et la reconnaissance précoce d’actions et de gestes, Sep. 2019, Eric Anquetil & Richard Kulpa
• PhD in progress (beginning September 2020): Pauline Morin, Adaptation des méthodes prédiction des efforts d’interaction pour l’analyse biomécanique du mouvement en milieu écologique, Ecole normale supérieure de Rennes, direction Georges Dumont et Charles Pontonnier
• PhD in progress (beginning June 2020): Lucas Mourot, Learning-Based Human Character Animation Synthesis for Content Production, Pierre Hellier (InterDigital), Ludovic Hoyet, Franck Multon, François Le Clerc (InterDigital).
• PhD in progress (beginning September 2020): Agathe Bilhaut, Stratégies perceptivo-motrices durant la locomotion des patients douloureux chroniques : nouvelles méthodes d’analyse et de suivi, Armel Crétual, Anne-Hélène Olivier, Mathieu Ménard (Institut Ostéopathie Rennes, M2S)
• PhD in progress (beginning October 2020): Emilie Leblong, Prise en compte des interactions sociales dans un simulateur de conduite de fauteuil roulant électrique en réalité virtuelle : favoriser l’apprentissage pour une mobilité inclusive, Anne-Hélène Olivier, Marie Babel (Rainbow team)
• PhD in progress (beginning November 2020): Thomas Chatagnon, Micro-to-macro energy-based interaction models for dense crowds behavioral simulations, Ecole normale supérieure de Rennes, Ludovic Hoyet, Anne-Hélène Olivier,Charles Pontonnier, Julien Pettré (Rainbow team).
• PhD in progress (beginning November 2020): Vicenzo Abichequer-Sangalli, Humains virtuels expressifs et réactifs pour la réalité virtuelle, Marc Christie, Ludovic Hoyet, Carol O'Sullivan (TCD, Ireland), Julien Pettré (Rainbow team).
• PhD in progress (beginning November 2020): Tairan Yin, Création de scènes peuplées dynamiques pour la réalité virtuelle, Marc Christie, Marie-Paule Cani (LIX), Ludovic Hoyet, Julien Pettré (Rainbow team).
• PhD in progress (beginning October 2020): Qian Li, Neural novel view synthesis of dynamic people from monocular videos, Adnane Boukhayma, Franck Multon.
• PhD in progress (beginning October 2021): Maé Mavromatis, Towards “Avatar-Friendly” Characterization of Virtual Reality Interaction Methods, Ferran Argelaguet (Hybrid team), Ludovic Hoyet, Anatole Lécuyer (Hybrid team).
• PhD in progress (beginning October 2021): Rebecca Crolan, Prediction of low back load during gymnastics landings for the prevention and follow-up of athlete injuries, Charles Pontonnier, Diane Haering, Matthieu Ménard (M2S Lab).
• PhD in progress (beginning November 2021): Rim Rekik, Learning and evaluating 3D human motion synthesis, Anne-Hélène Olivier, Stefanie Wuhrer (Morpheo team)

-

10.2.2 Juries

- -

• PhD defense: Univ. Claude Bernard Lyon1, Mingming Zhao, "Monitoring de la posture du conducteur", november 2021, Franck Multon, reviewer
• PhD defense: Univ. Montpellier, Carmela Calabrese, "Analysis of synchronization and leadership emergence in human group interaction", March 2021, Franck Multon, examiner
• HDR defense: Univ. Lille, Hazem Wannous, "Towards Understanding Human Behavior by Time-Series Analysis of 3D Motion", March 2021, Franck Multon, president
• PhD defense: université Paris Saclay, Benjamin Treussart, "Étude et conception d’un système de pilotage intuitif d’exosquelette pour l’assistance au port de charges", march 2021, Charles Pontonnier, reviewer
• PhD defense: Politecnico di Torino, Divyaksh Subhash Chander, "Modelling the Physical Human-Exoskeleton Interface", september 2021, Charles Pontonnier, examiner
• PhD defense: Institut Supérieur de l'Aéronautique et de l'Espace Toulouse, Rebai Soret, "Paradigme de Posner du laboratoire au monde réel : orientation de l’attention en espace avant et arrière ", November 2021, Anne-Hélène Olivier, reviewer.

International journals

• 11 articleS.Sheryl Bourgaize, M.Michael Cinelli, F.Florian Berton, B.Benjamin Niay, L.Ludovic Hoyet and A.-H.Anne-Hélène Olivier. Walking speed and trunk sway: Influence of an approaching person's gait pattern on collision avoidance.Journal of Vision219September 2021
  HALDOIback to text
• 12 articleH.Hugo Brument, G.Gerd Bruder, M.Maud Marchal, A.-H.Anne-Hélène Olivier and F.Ferran Argelaguet Sanz. Understanding, Modeling and Simulating Unintended Positional Drift during Repetitive Steering Navigation Tasks in Virtual Reality.IEEE Transactions on Visualization and Computer Graphics2711November 2021, 4300-4310
  HALDOIback to text
• 13 articleT.Thomas Chatagnon, A.-H.Anne-Hélène Olivier, L.Ludovic Hoyet, J.Julien Pettré and C.Charles Pontonnier. Modeling physical interactions in human crowds: a pilot study of individual response to controlled external pushes.Computer Methods in Biomechanics and Biomedical Engineering2021, 1-2
  HALback to text
• 14 articleB. C.Beatríz Cabrero Daniel, R.Ricardo Marques, L.Ludovic Hoyet, J.Julien Pettré and J.Josep Blat. A Perceptually-Validated Metric for Crowd Trajectory Quality Evaluation.Proceedings of the ACM on Computer Graphics and Interactive Techniques43September 2021, 1 - 18
  HALDOI
• 15 articleL.Louise Demestre, Y.Yohann Audoux, S.Stéphane Grange, N.Nicolas Bideau, G.Guillaume Nicolas, C.Charles Pontonnier and G.Georges Dumont. Using motion-based estimated action of the diver to characterize diving board dynamics: a pilot study.Computer Methods in Biomechanics and Biomedical Engineering2021, 1-2
  HALback to text
• 16 articleA.Aurelie Dommes, G.Gaetan Merlhiot, R.Régis Lobjois, N.-T.Nguyen-Thong Dang, F.Fabrice Vienne, J.Joris Boulo, A.-H.Anne-Hélène Olivier, A.Armel Crétual and V.Viola Cavallo. Young and older adult pedestrians' behavior when crossing a street in front of conventional and self-driving cars.Accident Analysis & Prevention159September 2021, 1-13
  HALDOIback to text
• 17 articleA.Abdallah Fayssoil, L. S.Lee S Nguyen, T.Tanya Stojkovic, H.Helene Prigent, R.Robert Carlier, H.Helge Amthor, J.Jean Bergounioux, J.Justine Zini, S.Sebastien Damez‐Fontaine, K.Karim Wahbi, P.Pascal Laforet, G.Guillaume Nicolas, A.Anthony Behin, G.Guillaume Bassez, F.France Leturcq, R.Rabah Ben Yaou, N.Nicolas Mansencal, D.Djillali Annane, F.Frédéric Lofaso and D.David Orlikowski. Determinants of diaphragm inspiratory motion, diaphragm thickening, and its performance for predicting respiratory restrictive pattern in Duchenne muscular dystrophy.Muscle & nerve. Supplement.651January 2022, 89-95
  HALDOI
• 18 articleR.Rebecca Fribourg, E.Evan Blanpied, L.Ludovic Hoyet, A.Anatole Lécuyer and F.Ferran Argelaguet. Does virtual threat harm VR experience?: Impact of threat occurrence and repeatability on virtual embodiment and threat response.Computers and Graphics100November 2021, 125-136
  HALDOIback to text
• 19 articleO.Olfa Haj Mahmoud, C.Charles Pontonnier, G.Georges Dumont, S.Stéphane Poli and F.Franck Multon. A neural networks approach to determine factors associated with self-reported discomfort in picking tasks.Human Factors2021, 1-24
  HALDOIback to text
• 20 articleS.Simon Hilt, T.Thomas Meunier, C.Charles Pontonnier and G.Georges Dumont. Biomechanical fidelity of simulated pick-and-place tasks: impact of visual and haptic renderings.IEEE Transactions on Haptics (ToH)2021, 8
  HALDOIback to text
• 21 articleC.Claire Livet, T.Théo Rouvier, G.Georges Dumont and C.Charles Pontonnier. An automatic and simplified approach to muscle path modelling.Journal of Biomechanical Engineering2021, 1-10
  HALback to textback to text
• 22 articleC.Claire Livet, T.Théo Rouvier, C.Christophe Sauret, G.Georges Dumont and C.Charles Pontonnier. Expected scapula orientation error regarding scapula-locator uncertainty while studying wheelchair locomotion.Computer Methods in Biomechanics and Biomedical Engineering2021, 1-2
  HALback to text
• 23 articleS.Sean Lynch, R.Richard Kulpa, L. A.Laurentius Antonius Meerhoff, A.Anthony Sorel, J.Julien Pettré and A.-H.Anne-Hélène Olivier. Influence of path curvature on collision avoidance behaviour between two walkers.Experimental Brain Research2391January 2021, 329-340
  HALDOIback to text
• 24 articleP.Pauline Morin, A.Antoine Muller, C.Charles Pontonnier and G.Georges Dumont. Foot contact detection through pressure insoles for the estimation of external forces and moments: application to running and walking.Computer Methods in Biomechanics and Biomedical Engineering2021, 1-2
  HALback to text
• 25 articleL.Lucas Mourot, L.Ludovic Hoyet, F.François Le Clerc, F.François Schnitzler and P.Pierre Hellier. A Survey on Deep Learning for Skeleton‐Based Human Animation.Computer Graphics ForumNovember 2021, 1-32
  HALDOI
• 26 articleC.Charles Pontonnier, S.Stéphane Marie, D. S.Divyaksh Subhash Chander, S.Simon Kirchhofer and M.Mathieu Gréau. Evaluating the action detection of an active exoskeleton: a muscle-control synchronicity approach for meat cutting assistance.Computer Methods in Biomechanics and Biomedical Engineering2021, 1-2
  HALback to text
• 27 articleC.Camille Pouliquen, G.Guillaume Nicolas, B.Benoit Bideau and N.Nicolas Bideau. Impact of Power Output on Muscle Activation and 3D Kinematics During an Incremental Test to Exhaustion in Professional Cyclists.Frontiers in Sports and Active Living22021, 1-11
  HALDOI
• 28 articleV.Victoria Rapos, M. E.Michael Eric Cinelli, R.Robyn Grunberg, S.Sheryl Bourgaize, A.Armel Crétual and A.-H.Anne-Hélène Olivier. Collision avoidance behaviours between older adult and young adult walkers.Gait and Posture88July 2021, 210-215
  HALDOIback to text
• 29 articleN.Natalie Snyder, M.Michael Cinelli, V.Victoria Rapos, A.Armel Crétual and A.-H.Anne-Hélène Olivier. Collision avoidance strategies between two athlete walkers: Understanding impaired avoidance behaviours in athletes with a previous concussion.Gait and Posture92November 2021, 24-29
  HALDOIback to text
• 30 articleX.Xi Wang, M.Marc Christie and E.Eric Marchand. Binary Graph Descriptor for Robust Relocalization on Heterogeneous Data.IEEE Robotics and Automation Letters2022
  HALback to textback to text

International peer-reviewed conferences

• 31 inproceedingsR.Robin Adili, B.Benjamin Niay, K.Katja Zibrek, A.-H.Anne-Hélène Olivier, J.Julien Pettré and L.Ludovic Hoyet. Perception of Motion Variations in Large-Scale Virtual Human Crowds.MIG 2021 - 14th Annual ACM SIGGRAPH Conference on Motion, Interaction and GamesVirtual Event Switzerland, FranceACMNovember 2021, 1-7
  HALDOIback to text
• 32 inproceedingsJ.Jean Basset, A.Adnane Boukhayma, S.Stefanie Wuhrer, F.Franck Multon and E.Edmond Boyer. Neural Human Deformation Transfer.3DV 2021 - 9th International Conference on 3D VisionLondres (on line event), United KingdomDecember 2021, 1-12
  HALback to text
• 33 inproceedingsH.Hugo Brument, M.Maud Marchal, A.-H.Anne-Hélène Olivier and F.Ferran Argelaguet Sanz. Studying the Influence of Translational and Rotational Motion on the Perception of Rotation Gains in Virtual Environments.SUI 2021 - Symposium on Spatial User InteractionVirtual Event, United StatesNovember 2021, 1-12
  HALDOIback to text
• 34 inproceedingsL.Ludovic Burg, C.Christophe Lino and M.Marc Christie. Real-Time Cinematic Tracking of Targets in Dynamic Environments.GI 2021 - Graphics Interface conferenceVancouver, CanadaMay 2021, 1-10
  HALback to text
• 35 inproceedingsD.Diane Dewez, L.Ludovic Hoyet, A.Anatole Lécuyer and F.Ferran Argelaguet Sanz. Towards “Avatar-Friendly” 3D Manipulation Techniques: Bridging the Gap Between Sense of Embodiment and Interaction in Virtual Reality.CHI 2021 - Conference on Human Factors in Computing SystemsYokohama, JapanACMMay 2021, 1-14
  HALDOIback to text
• 36 inproceedingsS.Shubhendu Jena, F.Franck Multon and A.Adnane Boukhayma. Monocular Human Shape and Pose with Dense Mesh-borne Local Image Features.FG 2021 - IEEE International Conference on Automatic Face and Gesture RecognitionJodhpur (online), IndiaDecember 2021, 1-5
  HALDOI
• 37 inproceedingsP.Pauline Morin, A.Antoine Muller, C.Charles Pontonnier and G.Georges Dumont. Studying the impact of internal and external forces minimization in a motion-based external forces and moments prediction method: application to fencing lunges.ISB 2021 - XXVIII Congress of the International Society of BiomechanicsStockholm, SwedenJuly 2021, 1
  HALback to text
• 38 inproceedingsI.Iana Podkosova, K.Katja Zibrek, J.Julien Pettré, L.Ludovic Hoyet and A.-H.Anne-Hélène Olivier. Exploring behaviour towards avatars and agents in immersive virtual environments with mixed-agency interactions.VR 2021 - 28th IEEE Conference on Virtual Reality and 3D User InterfacesLisbon, PortugalIEEE2021
  HALDOIback to text
• 39 inproceedingsP.Pierre Raimbaud, A.Alberto Jovane, K.Katja Zibrek, C.Claudio Pacchierotti, M.Marc Christie, L.Ludovic Hoyet, J.Julien Pettré and A.-H.Anne-Hélène Olivier. Reactive Virtual Agents: A Viewpoint-Driven Approach for Bodily Nonverbal Communication.IVA 2021: ACM International Conference on Intelligent Virtual AgentsVirtual Event Japan, FranceACMSeptember 2021, 164-166
  HALDOIback to text
• 40 inproceedingsN.Nitika Verma, A.Adnane Boukhayma, E.Edmond Boyer and J.Jakob Verbeek. Dual Mesh Convolutional Networks for Human Shape Correspondence.3DV 2021 - International Conference on 3D VisionSurrey, United KingdomDecember 2021, 1-10
  HALback to text
• 41 inproceedingsX.Xi Wang, M.Marc Christie and E.Eric Marchand. TT-SLAM: Dense Monocular SLAM for Planar Environments.ICRA 2021 - IEEE International Conference on Robotics and AutomationXi'an, ChinaMay 2021, 11690-11696
  HALback to text

National peer-reviewed Conferences

• 42 inproceedingsC.Charles Pontonnier, P.Pauline Morin, T.Théo Rouvier, L.Louise Demestre, E.Erwan Delhaye, C.Claire Livet, P.Pierre Puchaud, A.Antoine Muller, D.Diane Haering, C.Caroline Martin, A.Anthony Sorel, N.Nicolas Bideau, G.Guillaume Nicolas and G.Georges Dumont. CUSTOM: Une librairie pour l'analyse biomécanique du geste sportif.Conférence Sciences 2024Rennes, FranceMay 2021, 1-2
  HALback to text

Conferences without proceedings

• 43 inproceedingsR.Robin Courant, C.Christophe Lino, M.Marc Christie and V.Vicky Kalogeiton. High-Level Features for Movie Style Understanding.ICCV 2021 - Workshop on AI for Creative Video Editing and Understandingonline, FranceOctober 2021, 1-5
  HALback to text
• 44 inproceedingsL.Louise Demestre, Y.Yohann Audoux, S.Stéphane Grange, N.Nicolas Bideau, G.Guillaume Nicolas, C.Charles Pontonnier and G.Georges Dumont. Caractérisation d'un plongeoir à partir d'une méthode de prédiction des actions du plongeur: étude pilote.Conférence Sciences 2024Bruz, FranceMay 2021, 1-2
  HALback to text
• 45 inproceedingsL.Louise Demestre, F.François May, P.Pauline Morin, G.Guillaume Nicolas, N.Nicolas Bideau, G.Georges Dumont and C.Charles Pontonnier. Motion-based ground reaction forces and moments prediction method in a moving frame: a pilot study.ISB 2021 - XXVIII Congress of the International Society of BiomechanicsStockholm, SwedenJuly 2021, 1
  HALback to text
• 46 inproceedingsC.Claire Livet, T.Théo Rouvier, C.Charles Pontonnier and G.Georges Dumont. Open vs closed articular architecture of the forearm for an analysis of muscle recruitment during throwing motions.ISB 2021 - XXVIII Congress of the International Society of BiomechanicsStockholm, SwedenJuly 2021, 1
  HALback to text
• 47 inproceedingsW.William Mocaër, E.Eric Anquetil and R.Richard Kulpa. Online Spatio-Temporal 3D Convolutional Neural Network for Early Recognition of Handwritten Gestures.ICDAR 2021 - 16th International Conference on Document Analysis and RecognitionLausanne, SwitzerlandSeptember 2021, 221-236
  HALback to text

Doctoral dissertations and habilitation theses

• 48 thesisO.Olfa Haj Mahmoud. Discomfort estimation using machine learning for pick and place tasks in industry.Université Rennes2December 2021
  HAL

Reports & preprints

• 49 miscA.Alexandre Bruckert, M.Marc Christie and O.Olivier Le Meur. Where to look at the movies : Analyzing visual attention to understand movie editing.February 2021
  HALback to text
• 50 miscS.Shubhendu Jena, F.Franck Multon and A.Adnane Boukhayma. Monocular Human Shape and Pose with Dense Mesh-borne Local Image Features.December 2021
  HALback to text
• 51 miscP.Pratik Mullick, S.Sylvain Fontaine, C.Cécile Appert-Rolland, A.-H.Anne-Hélène Olivier, W. H.William H Warren and J.Julien Pettré. Analysis of emergent patterns in crossing flows of pedestrians reveals an invariant of 'stripe' formation in human data.December 2021
  HAL
• 52 miscN.Nicolas Olivier, K.Kelian Baert, F.Fabien Danieau, F.Franck Multon and Q.Quentin Avril. FaceTuneGAN: Face Autoencoder for Convolutional Expression Transfer Using Neural Generative Adversarial Networks.December 2021
  HALback to text
• 53 miscX.Xi Wang, M.Marc Christie and E.Eric Marchand. Supplementary Material: Binary Graph Descriptor for Robust Relocalization on Heterogeneous Data.November 2021
  HAL

Scientific popularization

• 54 miscA.-H.Anne-Hélène Olivier and J.Julien Pettré. Etude du comportement piéton en réalité virtuelle.July 2021
  HAL