Meta-learners for few-shot weakly-supervised medical image segmentation

Hugo N. Oliveira, Pedro H. T. Gama, Isabelle Bloch, Roberto Marcondes Cesar Jr.

Abstract: Highlights•Weakly-supervised segmentation via meta-learning with no ImageNet pretraining.•Proposal of pipelines to adapt classification learners to few-shot segmentation.•Experiments on medical images for weakly-supervised few-shot segmentation.•Performance analysis on multiple weakly-supervised scenarios and learning algorithms.•Code and demo will be made available to encourage reproducibility.