Andrews Cordolino Sobral
Ph.D. in Computer Vision and Machine Learning
Head of AI, ActiveEon, Paris, France
I am a computer vision specialist with over a decade of experience in research and industry, specializing in the intersection of Artificial Intelligence (AI), High-Performance Computing (HPC), and MLOps. As the Head of AI at ActiveEon, I lead a team of AI researchers to develop cutting-edge scalable AI workflows for distributed and parallel execution, including automated machine learning (AutoML) and hyperparameter optimization at scale.
Research Interests
• Computer Vision and Image Processing
• Machine Learning and Deep Learning
• Matrix and Tensor Methods for Data Decomposition
• Applied Mathematics and Optimization
Academic Background
• Ph.D. in Computer Vision and Pattern Recognition, University of La Rochelle, France
• Thesis: “Robust Low-Rank and Sparse Decomposition for Moving Object Detection: From Matrices to Tensors”
• M.Sc. in Mechatronics Engineering, Federal University of Bahia, Brazil
• B.Sc. in Computer Engineering, AREA1 Engineering School, Brazil
Professional Expertise
• End-to-end AI Workflow Development
• Large-scale Distributed AI Systems (multi-node, multi-GPU)
• Deep Learning Applications for Object Detection and Anomaly Detection
• Real-time Image and Video Processing on Embedded Systems (NVIDIA Jetson, Raspberry Pi)
Academic Contributions
I have authored several open-source libraries, including:
• BGSLibrary (C/C++) and LRSLibrary (MATLAB), tools for moving object detection in videos.
• Peer reviewer for journals like IEEE, Elsevier, and Springer in the fields of AI and Computer Vision.
• My publications and citation profile can be accessed on Google Scholar.
Programming and Tools
• Languages: Python, C++, MATLAB, Java, PHP
• Frameworks: PyTorch, TensorFlow, OpenCV, H2O, Tensor Toolbox
• Platforms: NVIDIA RAPIDS, Docker, HPC environments
Recognition
My research on robust data decomposition techniques has been featured in top-tier journals and conferences. I’ve delivered invited talks on AI and computer vision topics at global venues, including IPGG Paris and RFIA.
Publications
Matrix and tensor completion algorithms for background model initialization: A comparative evaluation
A. Sobral, E. Zahzah
Pattern Recognition Letters 2017
LRSLibrary: Low-Rank and Sparse tools for Background Modeling and Subtraction in Videos
A. Sobral, T. Bouwmans, E. Zahzah
OR-PCA with dynamic feature selection for robust background subtraction
S. Javed, A. Sobral, T. Bouwmans, Soon Ki Jung
ACM Symposium on Applied Computing 2015
OR-PCA with MRF for Robust Foreground Detection in Highly Dynamic Backgrounds
S. Javed, S. Oh, A. Sobral, T. Bouwmans, Soon Ki Jung
Asian Conference on Computer Vision 2014
BGS Library: A Library Framework for Algorithm’s Evaluation in Foreground/Background Segmentation
A. Sobral, T. Bouwmans
A comprehensive review of background subtraction algorithms evaluated with synthetic and real videos
A. Sobral, A. Vacavant
Computer Vision and Image Understanding 2014
Automated Mathematical Equation Structure Discovery for Visual Analysis
Caroline Silva, J. F. D. Souza, A. Vacavant, T. Bouwmans, A. Sobral
arXiv.org 2021
An automatic facial expression recognition system evaluated by different classifiers
Caroline Silva, A. Sobral, R. T. Vieira
Highway Traffic Congestion Classification using Holistic Properties
A. Sobral, Luciano Oliveira, L. Schnitman, Felippe de Souza
HIGHWAY TRAFFIC CONGESTION CLASSIFICATION
A. Sobral, Luciano Oliveira, L. Schnitman