Hidden Two-Stream Convolutional Networks for Action Recognition
This addresses the problem of slow and storage-intensive video action recognition for computer vision applications, representing a novel architectural improvement rather than a fundamental paradigm shift.
The paper tackles the computational inefficiency of two-stage action recognition systems that require pre-computed optical flow by introducing a hidden two-stream CNN architecture that implicitly captures motion information from raw video frames, achieving 10x faster processing while outperforming previous real-time approaches on four benchmark datasets.
Analyzing videos of human actions involves understanding the temporal relationships among video frames. State-of-the-art action recognition approaches rely on traditional optical flow estimation methods to pre-compute motion information for CNNs. Such a two-stage approach is computationally expensive, storage demanding, and not end-to-end trainable. In this paper, we present a novel CNN architecture that implicitly captures motion information between adjacent frames. We name our approach hidden two-stream CNNs because it only takes raw video frames as input and directly predicts action classes without explicitly computing optical flow. Our end-to-end approach is 10x faster than its two-stage baseline. Experimental results on four challenging action recognition datasets: UCF101, HMDB51, THUMOS14 and ActivityNet v1.2 show that our approach significantly outperforms the previous best real-time approaches.