Recognizing Dynamic Scenes with Deep Dual Descriptor based on Key Frames and Key Segments | Read Paper on Bytez