Can Unsupervised Segmentation Reduce Annotation Costs for Video Semantic Segmentation?

TL;DR

This paper explores reducing annotation costs in video semantic segmentation by leveraging unsupervised and coarse annotations with foundation models, achieving comparable performance with fewer manual labels.

Contribution

It demonstrates that using foundation models like SAM and SAM 2 with unannotated and coarse data can cut annotation efforts by a third without sacrificing accuracy.

Findings

Using SAM and SAM 2 automates mask generation for unannotated frames.

Dataset frame variety impacts performance more than sheer quantity.

Annotation effort can be reduced significantly with minimal performance loss.

Abstract

Present-day deep neural networks for video semantic segmentation require a large number of fine-grained pixel-level annotations to achieve the best possible results. Obtaining such annotations, however, is very expensive. On the other hand, raw, unannotated video frames are practically free to obtain. Similarly, coarse annotations, which do not require precise boundaries, are also much cheaper. This paper investigates approaches to reduce the annotation cost required for video segmentation datasets by utilising such resources. We show that using state-of-the-art segmentation foundation models, Segment Anything Model (SAM) and Segment Anything Model 2 (SAM 2), we can utilise both unannotated frames as well as coarse annotations to alleviate the effort required for manual annotation of video segmentation datasets by automating mask generation. Our investigation suggests that if used appropriately, we can reduce the need for annotation by a third with similar performance for video semantic segmentation. More significantly, our analysis suggests that the variety of frames in the dataset is more important than the number of frames for obtaining the best performance.