Coherent Temporal Synthesis for Incremental Action Segmentation

Guodong Ding, Hans Golong, Angela Yao; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 28485-28494

Data replay is a successful incremental learning technique for images. It prevents catastrophic forgetting by keeping a reservoir of previous data original or synthesized to ensure the model retains past knowledge while adapting to novel concepts. However its application in the video domain is rudimentary as it simply stores frame exemplars for action recognition. This paper presents the first exploration of video data replay techniques for incremental action segmentation focusing on action temporal modeling. We propose a Temporally Coherent Action (TCA) model which represents actions using a generative model instead of storing individual frames. The integration of a conditioning variable that captures temporal coherence allows our model to understand the evolution of action features over time. Therefore action segments generated by TCA for replay are diverse and temporally coherent. In a 10-task incremental setup on the Breakfast dataset our approach achieves significant increases in accuracy for up to 22% compared to the baselines.