1 paper · 1 filter
Yunheng Li, Zhongyu Li, Shanghua Gao +3
Effectively modeling discriminative spatio-temporal information is essential for segmenting activities in long action sequences. However, we observe that existing methods are limit…