Recent news
1/2023 [efficiency] PowerQuant: Automorphism Search for Non-Uniform Quantization accepted at ICLR 2023
11/2022 [internship] We are recruting @ ISIR and @ Datakalab. Please check our internship offers for 2023
10/2022 [efficiency] SPIQ: Data-Free Static Per-Channel Input Quantization accepted at WACV 2023
9/2022 [efficiency] SInGE: Sparsity via Integrated Gradients Estimation of Neuron Relevance accepted at NeurIPS 2022
9/2022 PhD defense of A. Pourchot with D. Mateus, C. Petitjean, S. Allassonnière, C. Wolf, A. Ducarouge, O. Sigaud. Congrats Alois!
8/2022 [medical imaging] Neural Architecture Search For Fracture Classification accepted at ICIP 2022
6/2022 PhD defense of A. Douillard with D. Larlus, D. Filliat, T. Tommasi, K. Alahari, M. Cord, T. Robert and K. Bailly (pres.)
6/2022 [efficiency] RED++: Data-Free Pruning of Deep Neural Networks via Input Splitting and Output Merging published in T-PAMI
5/2022 [face / multitask] Multi-Order Networks for Action Unit Detection published in IEEE Trans. Affect. Comput.
5/2022 [face / emotion] Privileged Attribution Constrained Deep Networks for Facial Expression Recognition accepted at ICPR 2022
4/2022 [efficiency] To Fold or Not to Fold: a Necessary and Sufficient Condition on Batch-Normalization Layers Folding [with code] accepted at IJCAI 2022.
4/2022 [NAS] An Extensive Appraisal of Weight-Sharing on the NAS-Bench-101 Benchmark published in Neurocomputing
3/2022 [medical] Artificial Intelligence: a systematic review of 20 years of contributions reported in Obstetrics and Gynecology journals published in JMIR
1/2022 [face / emotion] THIN: THrowable Information Networks and Application for Facial Expression Recognition In The Wild published in IEEE Trans. Affect. Comput.
9/2021 [efficiency] RED : Looking for Redundancies for Data-FreeStructured Compression of Deep Neural Networks accepted at NeurIPS 2021
Research interests
I am interested in machine learning, deep learning and computer vision with a particular focus on :
Efficient models using eg. pruning, quantification and distillation
Hybrid deep and ensemble learning
Multi-task learning
This research is mainly applied to real world applications :
Automatic Face Analysis (Emotion recognition, action unit detection and intensity estimation, facial landmark localisation and tracking)
Urban scene understanding (Object detection, people counting, re-identification, mask wearing analysis)
Medical applications (serious game for ASD children, dyspnea detection, ultrasound imaging)
Brief vitae
Since 2018 - Head of Research @ Datakalab (part time)
Since 2011 - Associate Professor @ SU
2019 - HDR (accreditation to supervise research) @ SU
2010/2011 - Postdoc @ Telecom Paris
2010 - PhD in computer science @ UPMC (now SU)
Contact
Institut des Systèmes Intelligents et de Robotique
Campus Pierre et Marie Curie
Pyramide – Tour 65-66 3ème étage
4, place Jussieu 75005 Paris – France
Tel: +33 (0) 1 44 27 63 85
e-mail: firstname.lastname (at) sorbonne-universite.fr