1 paper · 1 filter
Yao Tian, Le Yang, Binglu Wang
Vision cues are available and informative for pedestrian action prediction, but obtaining stable target-centric representations from video frames remains challenging without frame-…