Jethro's Braindump

chen_big_2020: Big Self-Supervised Models are Strong Semi-Supervised Learners

Big Self-Supervised Models are Strong Semi-Supervised Learners

Key Idea

The paper proposes a three-step semi-supervised learning framework for visual data

  1. unsupervised or self-supervised pretraining using a modified SimCLR

  2. Supervised fine-tuning

  3. Distillation using unlabeled data

Links to this note