Biography
Bogdan Stanciulescu is a researcher whose work focuses on the analysis and interpretation of visual and spatiotemporal data, with major applications in the fields of computer vision, robotics, and autonomous systems. His research addresses issues such as action detection in sports videos, particularly in soccer, where he explores the integration of contextual models and graph neural networks to improve prediction accuracy by combining visual information with the state of the game. His expertise also extends to localization and 3D reconstruction, with significant contributions to the use of Neural Radial Fields (NeRF) for camera relocalization and the synthesis of new views, as well as to the development of efficient visual localization methods for autonomous vehicles. In addition, he has worked on multi-agent trajectory prediction in urban environments, proposing innovative architectures such as temporal graph neural networks to model the complex interactions between agents and road infrastructure. His recent work also includes applications in hyperspectral medical imaging, where he has contributed to the development of classification methods to assist in brain surgery.
Publication(s)
-
2026
Beyond Pixels: Leveraging the Language of Soccer to Improve Spatio-Temporal Action Detection in Broadcast Videos DOI : 10.1007/978-3-032-07343-3_44
-
2025
Game State and Spatio-Temporal Action Detection in Soccer Using Graph Neural Networks and 3D Convolutional Networks DOI : 10.5220/0013161100003905
-
2023
CROSSFIRE: Camera Relocalization On Self-Supervised Features from an Implicit Representation DOI : 10.1109/ICCV51070.2023.00030
-
2023
ImPosing: Implicit Pose Encoding for Efficient Visual Localization DOI : 10.1109/WACV56688.2023.00291
-
2023
TSGN: Temporal Scene Graph Neural Networks with Projected Vectorized Representation for Multi-Agent Motion Prediction DOI : 10.1109/IV55152.2023.10186764
-
2022
THOMAS: TRAJECTORY HEATMAP OUTPUT WITH LEARNED MULTI-AGENT SAMPLING
-
2022
GOHOME: Graph-Oriented Heatmap Output for future Motion Estimation DOI : 10.1109/ICRA46639.2022.9812253
-
2022
Assessing Cross-dataset Generalization of Pedestrian Crossing Predictors DOI : 10.1109/IV51971.2022.9827083
-
2022
CoordiNet: Uncertainty-aware pose regressor for reliable vehicle localization DOI : 10.1109/WACV51458.2022.00191
-
2021
HOME: Heatmap Output for future Motion Estimation DOI : 10.1109/ITSC48978.2021.9564944
-
2021
LENS: Localization enhanced by NeRF synthesis
-
2020
Predicting intentions of pedestrians from 2d skeletal pose sequences with a representation-focused multi-branch deep learning network DOI : 10.3390/a13120331
-
2019
In-Vivo Hyperspectral Human Brain Image Database for Brain Cancer Detection DOI : 10.1109/ACCESS.2019.2904788
-
2018
Wheel-rail contact analysis system using spectral signatures for train automation and traffic management DOI : 10.1109/IRIS.2017.8250146
-
2018
Spatio-spectral classification of hyperspectral images for brain cancer detection during surgical operations DOI : 10.1371/journal.pone.0193721
-
2018
An intraoperative visualization system using hyperspectral imaging to aid in brain tumor delineation DOI : 10.3390/s18020430
-
2017
Pollution detection on rail surface for adhesion evaluation using multispectral images DOI : 10.1109/DICTA.2017.8227477
-
2017
Real-Time method for general road segmentation DOI : 10.1109/IVS.2017.7995758
-
2017
Pollution discrimination on rail surface for adhesion evaluation using hyperspectral signatures DOI : 10.1109/ITSC.2017.8317641
-
2016
Unsupervised clustering of hyperspectral images of brain tissues by hierarchical non-negative matrix factorization
-
2015
Person re-identification using the silhouette shape described by a point distribution model DOI : 10.1109/WACV.2015.128
-
2015
Clustering and classifying deformations for Shape Regression applied to the human body DOI : 10.1109/ISPA.2015.7306027
-
2014
Hybrid visual and inertial RANSAC for real-time motion estimation DOI : 10.1109/ICIP.2014.7025035
-
2014
General road detection algorithm: A computational improvement DOI : 10.5220/0004935208250830
-
2014
Real-time traffic sign recognition in three stages DOI : 10.1016/j.robot.2012.07.019
-
2013
Rail and turnout detection using gradient information and template matching DOI : 10.1109/ICIRT.2013.6696299
-
2013
Turnout detection and classification using a modified HOG and template matching DOI : 10.1109/ITSC.2013.6728530
-
2012
Real-time traffic-sign recognition using tree classifiers DOI : 10.1109/TITS.2012.2225618
-
2012
Rail extraction technique using gradient information and a priori shape model DOI : 10.1109/ITSC.2012.6338870
-
2011
Warning traffic sign recognition using a HOG-based K-d tree DOI : 10.1109/IVS.2011.5940454
-
2011
Segmentation masks for real-time traffic sign recognition using weighted HOG-based trees DOI : 10.1109/ITSC.2011.6082953
-
2011
Real-time traffic sign recognition using spatially weighted HOG trees DOI : 10.1109/ICAR.2011.6088571
-
2011
Traffic sign classification using K-d trees and random forests DOI : 10.1109/IJCNN.2011.6033494
-
2010
Classifying bags of keypoints using HMMs DOI : 10.1109/AICCSA.2010.5587039
-
2008
Person re-identification in multi-camera system by signature based on interest point descriptors collected on short video sequences DOI : 10.1109/ICDSC.2008.4635689
-
2006
Combining adaBoost with a hill-climbing evolutionary feature search for efficient training of performant visual object detectors DOI : 10.1142/9789812774118_0104
-
2003
Physical modeling framework for robotics applications
Teaching
Course 1, 2, or 3 - (PI IDS)
Touchdown/Touchdown (Pass Interference IDS)
Project (PI IDS)
Applied Mathematics: Robotics, Computer Vision, and Control Systems (MAREVA) track
Second-Year Curriculum The courses and educational activities offered in the second year are the same for all students who have chosen the MAREVA program and take place in Paris. The second-year curriculum focuses on the study of Complex Systems: discrete-time dynamic systems, filtering and identification, embedded sensors and data fusion, image processing, and medical imaging. This module is supplemented by introductory lectures on robotics and ITS, virtual reality and augmented reality, C++ programming and graphical programming sessions, as well as mini-project sessions conducted in pairs, supervised by researchers from the centers, and applying the concepts covered in class. Here are a few examples of mini-projects completed in recent years: lateral control (Lane Keeping, ESP) and longitudinal control of a vehicle (ACC); color-sensor-based control of a small car on a white line; evaluation of a consumer motion-capture system (the Wiimote); 3D point cloud registration for digital mapping; modeling and stabilization of a diver; image editing based on hierarchical segmentation; modeling and control of construction cranes; estimation and control algorithms for a mobile robot developed by SAGEM; development of parking algorithms, validation, and testing using 3D graphics software; implementation of human-environment interactions in a virtual kitchen. Third-Year Curriculum In the third year, the activities for the month of October remain the same for all students. The first two weeks in Paris are devoted to courses, notably on computer vision and image processing in the automotive context, on nonlinear control and its applications in robotics, on humanoid robots, on discrete-event systems, and on systems with delays.... A one-week intensive course in computer vision and mathematical morphology held in November rounds out the training for the Computer Vision and Robotics specializations. The last two weeks of October are dedicated to mini-projects (at a more advanced level than those in the second year), as well as to the optional field trip during which students visit laboratories and companies. Over the past four years, the trip has been organized to Italy and the Rhône-Alpes region (STMicroelectronics, ISPRA, INRIA Rhône-Alpes, LAG ...), in the Nice and Monaco region (INRIA Sophia-Antipolis, Thalès Alénia Space, the Automata Museum ...), in Germany in Stuttgart and Munich (Bosch, Mercedes, German Aerospace ...), and in Toulouse (CNES, LAAS, Pierre Fabre Laboratories...), and in Bordeaux (LABRI, LAPS, Laser MégaJoule, Thales, EvTronic, BeTomorrow...). These trips provide excellent opportunities to establish valuable contacts for elective internships in the third year. Internship topics for the third year are pooled and proposed by faculty members from the school’s various applied mathematics centers (CAOR, CAS, CMA, CMM), often in connection with their collaborative work with industry partners. It should be noted that each year, the number of internships offered far exceeds the number of students taking the elective! Some representative elective topics covered in recent years: Control Systems: Exo-atmospheric guidance of Ariane V – EADS (Control Systems); Signal processing using filter banks (IFP); Development of a method for analyzing ECG signals using inverse scattering (INRIA); Simulation and control of humanoid robots (ATR Computational Neuroscience Laboratory, Kyoto); Synthesis of flight control correctors for a supersonic aerobatic vehicle (ONERA); Validation of vehicle models for control and command (PSA) Quantification of maximum allowable disturbances on the control surfaces of a winged vehicle during atmospheric reentry (EADS) Parameter estimation for quantum systems (Princeton University) Optimization of electricity generation (EDF, Clamart) Sizing of natural gas transmission networks, (GdF Suez, Saint-Denis) Robotics: Computer Vision-Assisted Driving (INRIA/CAOR) Improving Road Safety and Comfort via Global Chassis Control (PSA) Identification of Available Traction for a Vehicle (NEXYAD, ARCOS project) Configuration of a force-feedback system in virtual reality (CEA/CAOR) Vehicle-to-infrastructure communication on highways (ASFA) Virtual vehicle coupling for convoy driving (INRIA) Development of a humanoid robot behavior editor (Aldebaran Robotics) Motion planning for humanoid robots (Joint Research Laboratory Tsukuba) Development of a social robot (Advanced Telecommunications Research, Kyoto) The Cybercars Olympics: a competition featuring smart vehicles on the “road of the future” in Saint-Brieuc (Côtes d’Armor General Council) Simulation and position estimation for a quadcopter drone (Parrot, Paris) Preparation of the French pavilion at the Shanghai World Expo on ITS Road traffic modeling, Millennium project (University of Berkeley) Neural algorithms for perception (MIT, Cambridge) Spherical vision sensor for autonomous mobile robotics (INRIA Sophia-Antipolis) Vision: Automatic image annotation (LTU, Paris) Indexing of medical images (CEA, Fontenay-aux-Roses) Environmental sensor for a telematics service (PSA, Vélizy) Segmentation of medical images for radiation therapy (Gustave Roussy Institute, Villejuif) Fingerprint image analysis (SAGEM, Eragny-sur-Oise) Analysis of sports video sequences (Thomson Broadcast, Breda, Netherlands) Compression/deconvolution optimization (Alcatel Space Industries, Cannes) Quantitative cytology and drug discovery (CSIRO, Sydney, Australia) 3D morphogenesis of the mouse kidney (Monash University, Clayton, Australia) Video scene interpretation (Bosch, Hildesheim, Germany) Geolocation of a landscape (Google Earth, USA) Quality control of photovoltaic glass (Saint-Gobain, China) Image analysis for studying the mechanical properties of bone (University of Adelaide, Australia) Special features of the track: Students greatly appreciate working on these mini-projects. It is an excellent way to deepen their understanding of and apply the concepts covered in class, as well as a great opportunity to meet researchers from the school’s various laboratories and gain their first—albeit limited—experience in the research environment.
PhD supervision
- 2025 Semantic perception through spatio-spectral analysis of the scene. IVANOVA Ivanina
- 2023 Active perception for night scene understanding through vehicle lighting DE MOREAU Simon
- 2023 Sports action recognition algorithms with applications in football games OCHIN Jérémie
- 2022 Real-time 3D visualization through neural rendering ANDO Angelika
- 2021 Sports action recognition algorithms with applications in football games BABANA Mohamed
- 2020 Smart prediction of vehicle trajectories in various autonomous driving scenarios GILLES Thomas
- 2020 Machine learning algorithms for real-time vehicle visual localization MOREAU Arthur
- 2019 Analysis of pedestrian movements and gestures via embedded camera for predicting their intentions GESNOUIN Joseph