Self-training is a new semi-supervised learning approach that works very well on many computer vision task. Here we explain how it works and the last experimentations that have been done in the tasks of object detection and semantic segmentation.
The main paper explained in the video is: "Rethinking Pre-training and Self-training" https://arxiv.org/abs/2006.06882
The paper that introduces the Noisy Student "Self-training with Noisy Student improves ImageNet classification" can be found here: https://arxiv.org/abs/1911.04252
