This talk focuses on weakly supervised object detection, which aims to learn to predict bounding boxes from only image-level labels. The talk provides insights from 50+ papers in the literature; provides a common framework for weakly supervised object detection methods; finally discusses the recent state-of-the-art papers.

More videos from the same ECCV tutorial:

Tutorial website:
https://hbilen.github.io/wsl-eccv20.github.io

Slides pdf:
https://github.com/hbilen/wsl-eccv20.github.io/blob/master/slides/2-Tutorial%20on%20Weakly%20Supervised%20Object%20Detection.pdf

source