1 paper
Tieqiao Wang, Sinisa Todorovic
Most recent work on action segmentation relies on pre-computed frame features from models trained on other tasks and typically focuses on framewise encoding and labeling without ex…