D3: Recognizing dynamic scenes with deep dual descriptor based on key frames and key segments

Sungeun Hong, Jongbin Ryu, Woobin Im, Hyun Seung Yang