Top-down Attention Recurrent VLAD Encoding for Action Recognition in Videos
This work addresses action recognition for video analysis, but it is incremental as it builds on existing VLAD encoding and attention methods.
The paper tackles action recognition in videos by proposing TA-VLAD, a deep recurrent architecture with spatial attention that uses class-specific activation maps to weight features before encoding, achieving state-of-the-art accuracy on HMDB51 and UCF101 benchmarks.
Most recent approaches for action recognition from video leverage deep architectures to encode the video clip into a fixed length representation vector that is then used for classification. For this to be successful, the network must be capable of suppressing irrelevant scene background and extract the representation from the most discriminative part of the video. Our contribution builds on the observation that spatio-temporal patterns characterizing actions in videos are highly correlated with objects and their location in the video. We propose Top-down Attention Action VLAD (TA-VLAD), a deep recurrent architecture with built-in spatial attention that performs temporally aggregated VLAD encoding for action recognition from videos. We adopt a top-down approach of attention, by using class specific activation maps obtained from a deep CNN pre-trained for image classification, to weight appearance features before encoding them into a fixed-length video descriptor using Gated Recurrent Units. Our method achieves state of the art recognition accuracy on HMDB51 and UCF101 benchmarks.