Adversarial Cross-Domain Action Recognition with Co-Attention
It addresses the problem of temporal misalignment in cross-domain action recognition for video analysis, which is an incremental advancement over existing image-based techniques.
The paper tackles cross-domain action recognition by proposing a Temporal Co-attention Network (TCoN) that aligns temporal features between domains, resulting in significant improvements over previous methods on three datasets.
Action recognition has been a widely studied topic with a heavy focus on supervised learning involving sufficient labeled videos. However, the problem of cross-domain action recognition, where training and testing videos are drawn from different underlying distributions, remains largely under-explored. Previous methods directly employ techniques for cross-domain image recognition, which tend to suffer from the severe temporal misalignment problem. This paper proposes a Temporal Co-attention Network (TCoN), which matches the distributions of temporally aligned action features between source and target domains using a novel cross-domain co-attention mechanism. Experimental results on three cross-domain action recognition datasets demonstrate that TCoN improves both previous single-domain and cross-domain methods significantly under the cross-domain setting.