1 paper
Kamel Aouaidjia, Wenhao Zhang, Aofan Li +1
Existing supervised action segmentation methods depend on the quality of frame-wise classification using attention mechanisms or temporal convolutions to capture temporal dependenc…