
This research paper introduces a novel framework for online temporal action segmentation (TAS) in videos, featuring an adaptive memory bank to capture temporal context and a context-aware feature augmentation module to enhance frame representations, leading to state-of-the-art performance in online action segmentation.


coremsg

online-temporal-action-segmentation-with-adaptive-memory-and-context-aware-feature-augmentation


Online Temporal Action Segmentation with Adaptive Memory and Context-Aware Feature Augmentation


title_rewrite


Introducing two methods, surround dense sampling and Online Temporally Aware Label Cleaning (O-TALC), to improve the performance of online temporal action segmentation by addressing the issues of inaccurate segment boundaries and oversegmentation.


online-temporal-action-segmentation-with-surround-sampling-and-temporally-aware-label-cleaning


Online Temporal Action Segmentation with Surround Sampling and Temporally Aware Label Cleaning
