A5008408311- Domain Adaptation and Few-Shot Learning
- Motor Control and Adaptation
- Sports Performance and Training
- Face recognition and analysis
- Sport Psychology and Performance
- Generative Adversarial Networks and Image Synthesis
- Anomaly Detection Techniques and Applications
- Face and Expression Recognition
- Human Pose and Action Recognition
- Cancer-related molecular mechanisms research
- Neural Networks and Applications
- Winter Sports Injuries and Performance
- Multimodal Machine Learning Applications
- Action Observation and Synchronization
- Balance, Gait, and Falls Prevention
- Computer Graphics and Visualization Techniques
- Radiomics and Machine Learning in Medical Imaging
- Seismic Imaging and Inversion Techniques
- COVID-19 diagnosis using AI
- Cell Image Analysis Techniques
- Image and Signal Denoising Methods
- Machine Learning and Data Classification
- Advanced Image Processing Techniques
- Data Mining Algorithms and Applications
- Sports Analytics and Performance
Normandie Université
2015-2025
GREYC
2021-2025
Université de Caen Normandie
2024-2025
École Nationale Supérieure d'Ingénieurs de Caen
2024-2025
Centre National de la Recherche Scientifique
2024-2025
Laboratoire d'Informatique, du Traitement de l'Information et des Systèmes
2014-2023
Université de Rouen Normandie
2013-2023
Institut National des Sciences Appliquées Rouen Normandie
2013-2023
Université de Technologie de Compiègne
2007-2008
Probabilistic inversion within a multiple-point statistics framework is often computationally prohibitive for high-dimensional problems. To partly address this, we introduce and evaluate new training-image based approach complex geologic media. Our relies on deep neural network of the generative adversarial (GAN) type. After training using image (TI), our proposed spatial GAN (SGAN) can quickly generate 2D 3D unconditional realizations. A key characteristic SGAN that it defines (very)...
This study investigated the functional intra-individual movement variability of ice climbers differing in skill level to understand how icefall properties were used by participants as affordances adapt inter-limb coordination patterns during performance. Seven expert and seven beginners observed they climbed a 30 m icefall. Movement positioning left right hand tools, crampons climber’s pelvis over first 20 climb recorded digitized using video footage from camera (25 Hz) located perpendicular...
Contrastive representation learning has proven to be an effective self-supervised method. Most successful approaches are based on Noise Estimation (NCE) and use different views of instance as positives that should contrasted with other instances, called negatives, considered noise. However, several instances in a dataset drawn from the same distribution share underlying semantic information. A good data contain relations, or similarity, between instances. implicitly learns relations but...
This study investigated a new performance indicator to assess climbing fluency (smoothness of the hip trajectory and orientation climber using normalized jerk coefficients) explore effects practice hold design on performance. Eight experienced climbers completed four repetitions two, 10-m high routes with similar difficulty levels, but varying in graspability (holds one edge vs holds two edges). An inertial measurement unit was attached hips each collect 3D acceleration data compute...
We present COMEDIAN, a novel pipeline to initialize spatiotemporal transformers for action spotting, which involves self-supervised learning and knowledge distillation. Action spotting is timestamp-level temporal detection task. Our consists of three steps, with two initialization stages. First, we perform spatial transformer using short videos as input. Additionally, that enhances the transformer's outputs global context through distillation from pre-computed feature bank aligned each video...
The aim of this study was to investigate how the affordances an indoor climbing wall changed for intermediate climbers following a period practice during which hold orientation manipulated within learning and transfer protocol. protocol consisted four sessions, in eight randomly ascended three different routes fixed absolute difficulty (5c on French scale), as fluently possible. All were 10.3 m height composed 20 hand-holds at same locations artificial wall; only orientations altered: (i)...
This paper presents a novel application of machine learning method to automatically detect and classify climbing activities using inertial measurement units (IMUs) attached the wrists, feet, pelvis climber. detection/classification can be useful for research in sport science replace manual annotation where IMUs are becoming common. Detection requires phase with construct statistical models. Full-body activity is then classified based on detection each IMU.
The aim of this study is to compare and validate an Inertial Measurement Unit (IMU) relative optic system, propose methods for pattern recognition capture behavioural dynamics during sport performance. IMU validation was conducted by comparing the motions two arms a compass, which equipped with IMUs reflective landmarks detected multi-camera system. Spearman's rank correlation tests showed good correlations between especially when angles were normalized. Bland-Altman plot, root mean square...
Abstract Contrastive representation learning has proven to be an effective self-supervised method for images and videos. Most successful approaches are based on Noise Estimation (NCE) use different views of instance as positives that should contrasted with other instances, called negatives, considered noise. However, several instances in a dataset drawn from the same distribution share underlying semantic information. A good data contain relations between or similarity dissimilarity,...
Research background and hypothesis. Ice climbers determine their own ascent paths by creating holes with crampons ice tools. The coupling of upper lower limbs thus emerges from the icefall environment without prescriptions for one mode coordination. aim. aim this study was to analyse / limb coordination different skill level explore how environmental constraint (ice fall shape) is used adapt motor behaviour.Research methods. Six elite fi ve beginners climbed a 30m icefall, respectively in...
The aim of this study was to propose a method for full-body movement pattern recognition in climbing, by computing the 3D unitary vector four limbs and pelvis during performance. One climber with an intermediate skill level traversed two easy routes similar rates difficulty (5c on French scale), 10m height under top-rope conditions. first route simply designed allow horizontal edge-hold grasping, while second more complexity both vertical grasping. Five inertial measurement units (IMUs) were...