I’m currently a PhD student at Sorbonne University under the supervision of
Professor Matthieu Cord. My thesis is on incremental learning. I’m also working
part-time as a Research Scientist at Heuritech, deploying convolution nets for fashion.
PLOP: Learning without Forgetting for Continual Semantic Segmentation
Arthur Douillard, Yifu Chen, Arnaud Dapogny, Matthieu Cord
Continual Semantic Segmentation with Local POD, a multi-scale distillation scheme preserving long- and short-range spatial relationships at feature level, and an uncertainty-based pseudo-labeling handling background shift.
Arthur Douillard, Eduardo Valle, Charles Ollion, Thomas Robert, Matthieu Cord
Instead of optimizing the past and current tasks, we train a model to be good on all tasks, future included. The incorporation of future tasks with ZeroShot allows us to gain insights useful to minimize interference with old and …