Action Segmentation with Joint Self-Supervised Temporal Domain Adaptation
This work addresses the challenge of domain discrepancy in action segmentation for video analysis, offering a novel adaptation method that reduces reliance on labeled data, though it is incremental in improving existing domain adaptation techniques.
The paper tackles the problem of spatiotemporal variations in action segmentation by proposing a self-supervised temporal domain adaptation method, which outperforms state-of-the-art approaches on benchmark datasets with significant improvements, such as increasing F1@25 scores from 59.6% to 69.1% on Breakfast, and requires only 65% of labeled data for comparable performance.
Despite the recent progress of fully-supervised action segmentation techniques, the performance is still not fully satisfactory. One main challenge is the problem of spatiotemporal variations (e.g. different people may perform the same activity in various ways). Therefore, we exploit unlabeled videos to address this problem by reformulating the action segmentation task as a cross-domain problem with domain discrepancy caused by spatio-temporal variations. To reduce the discrepancy, we propose Self-Supervised Temporal Domain Adaptation (SSTDA), which contains two self-supervised auxiliary tasks (binary and sequential domain prediction) to jointly align cross-domain feature spaces embedded with local and global temporal dynamics, achieving better performance than other Domain Adaptation (DA) approaches. On three challenging benchmark datasets (GTEA, 50Salads, and Breakfast), SSTDA outperforms the current state-of-the-art method by large margins (e.g. for the F1@25 score, from 59.6% to 69.1% on Breakfast, from 73.4% to 81.5% on 50Salads, and from 83.6% to 89.1% on GTEA), and requires only 65% of the labeled training data for comparable performance, demonstrating the usefulness of adapting to unlabeled target videos across variations. The source code is available at https://github.com/cmhungsteve/SSTDA.