LIVIA, the Laboratory of Imaging, Vision and Artificial Intelligence, builds cost-effective deep-learning models from weakly-annotated data, for healthcare, security, industry and beyond.
Our expertise spans deep and machine learning, computer vision, pattern recognition and the optimization of complex systems, with a focus on adaptive, weakly-supervised and multimodal learning for deployable models.
Adaptive, incremental, weakly-supervised and multimodal learning for deep models under limited annotations.
Visual perception of 2D and 3D scenes; static and dynamic modelling of environments.
Biometrics, classifier ensembles and dynamic selection for robust decision-making.
Models that adapt to users and shifting conditions in real-time, resource-limited settings.
Multimodal fusion of vision, audio, language and physiological signals.
Convex and combinatorial optimization, evolutionary and multi-objective methods.
From hospitals to smart buildings, LIVIA turns weakly-annotated data into reliable systems across six core domains.
Medical and satellite imaging.
Biometrics and surveillance.
Monitoring health and marketing.
Industrial vision and secure embedded AI.
Document security and explainable AI.
Natural language understanding.
Ranked among Canada's leading vision & learning groups, with strong funding, training and publication output.
We release large-scale, carefully annotated datasets, with code, pretrained weights and reproducible benchmarks.
The first multimodal, subject-based video dataset for recognizing ambivalence and hesitancy (A/H) for digital behavioural-change interventions.
A large-scale dataset for single-image super-resolution in confocal fluorescence microscopy, with real low-resolution images captured by the microscope.
LIVIA is embedded in leading national and international research networks.
International Laboratory on Learning Systems.
Regroupement pour l'étude des environnements partagés intelligents répartis.
Institut de valorisation des données.
Affiliate, Montreal Behavioural Medicine Centre.
Associate member, Goodman Cancer Institute.
Collaborations in medical imaging and health.
★ The 3rd AH Video Recognition Challenge is open at ABAW 11th, ECCV 2026, built on our BAH dataset. Registration is open now; the test set releases Jul 10, 2026.
LIVIA presents 5 papers at ICCV 2024, point-cloud adaptation, facial-expression capture, few-shot learning, vision-language models and object detection.
🎉 LIVIA presents 7 papers at CVPR 2024, conformal prediction, anomaly detection, test-time adaptation, point-cloud processing and more.
We are looking for motivated PhD students, postdocs and Master's students to help shape the future of computer vision and AI.