Author(s)

Wenrao Pang

Corresponding Author

Wenrao Pang

ABSTRACT

Weakly supervised learning are challenging tasks in video segmentation. A survey on deep learning technique for video segmentation from all angles has been published in [1], including 3 major categories and 7 specific small problem directions. However, the literature only describe the main algorithms from 2017 - July 2021. This article mainly reviews the task of video segmentation from the perspective of weak supervision for video segmentation. Besides, we show the main data sets usually used in this assignment and analysis the main algorithms and the latest ones that not mentioned in above survey. Additionally, we state a famous challenge in this field and the evaluation methods to demonstrate the performance of algorithms.

KEYWORDS

Weakly supervised learning, video segmentation, semi-supervised learning