Selected Publications

Model Ratatouille: Recycling Diverse Models for Out-of-Distribution Generalization

arXiv
We propose a new fine-tuning strategy that improves OOD generalization in computer vision by recycling and averaging weights specialized on diverse auxiliary tasks.
PDF Code

Diverse Weight Averaging for Out-of-Distribution Generalization

NeurIPS 2022
To improve out-of-distribution generalization on DomainBed, we average diverse weights obtained from different training runs; this strategy is motivated by an extension of the bias-variance theory to weight averaging.
PDF Code

DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion

CVPR 2022
We propose a new dynamic transformer architecture for continual learning with state-of-the-art performances.
PDF Code

Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization

ICML 2022
We introduce and motivate a new regularization that enforces invariance in the domain-level gradient variances across the different training domains in order to improve out-of-distribution generalization.
PDF Code Poster Slides

MixMo: Mixing Multiple Inputs for Multiple Outputs via Deep Subnetworks

ICCV 2021
We introduce a new generalized framework for learning multi-input multi-output subnetworks and study how to best mix the inputs. We obtain sota on CIFAR and Tiny ImageNet by better leveraging the expressiveness of large networks.
PDF Code Poster Slides

DICE: Diversity in Deep Ensembles via Conditional Redundancy Adversarial Estimation

ICLR 2021
Driven by arguments from information theory, we introduce a new learning strategy for deep ensembles that increases diversity among members: we adversarially prevent features from being conditionally redundant.
PDF Poster Slides Video OpenReview

CORE: Color Regression for Multiple Colors Fashion Garments

CVPR 2022 CVFAD Workshop
We detect continuous colors for fashion garments using a new architecture.
PDF Poster

OMNIA Faster R-CNN: Detection in the Wild through Dataset Merging and Soft Distillation

arXiv Preprint
We improve performances of object detectors via combining different datasets through soft distillation.
PDF Video

Leveraging Weakly Annotated Data for Fashion Image Retrieval and Label Prediction

ICCV 2017 Fashion Workshop
We present a method to learn a visual representation adapted for e-commerce products.
PDF

Talks/Posts

PRAIRIE Artificial Intelligence Summer School: Key Takeaways   Medium
July 2018  ·   Grenoble
Semi-supervised Learning for Multilingual Sentence Representation   PDF   Video
March 2016  ·   Sorbonne / Heuritech

Teaching

Deep Learning for Computer Vision
Fall 2020/ Fall 2021  ·   Sorbonne University  ·   Teacher Assistant
Deep Learning
Fall 2017/Fall 2018  ·   Master Data Science l'X-Paris Saclay  ·   Teacher Assistant
Mathematics
Nov 2011 - Mar 2012  ·   Fondation d'Auteuil  ·   Volunteer Teacher and Youth Leader