Temporal Context Aggregation for Video Retrieval with Contrastive Learning
This addresses the need for efficient and accurate video retrieval systems, offering a novel method for handling temporal dependencies, though it is incremental in improving existing contrastive learning approaches.
The paper tackles the problem of modeling long-range semantic dependencies in video retrieval by proposing TCA, a framework that uses self-attention for temporal context aggregation and supervised contrastive learning with hard negative mining, achieving a ~17% mAP improvement on FIVR-200K and 22x faster inference compared to frame-level features.
The current research focus on Content-Based Video Retrieval requires higher-level video representation describing the long-range semantic dependencies of relevant incidents, events, etc. However, existing methods commonly process the frames of a video as individual images or short clips, making the modeling of long-range semantic dependencies difficult. In this paper, we propose TCA (Temporal Context Aggregation for Video Retrieval), a video representation learning framework that incorporates long-range temporal information between frame-level features using the self-attention mechanism. To train it on video retrieval datasets, we propose a supervised contrastive learning method that performs automatic hard negative mining and utilizes the memory bank mechanism to increase the capacity of negative samples. Extensive experiments are conducted on multiple video retrieval tasks, such as CC_WEB_VIDEO, FIVR-200K, and EVVE. The proposed method shows a significant performance advantage (~17% mAP on FIVR-200K) over state-of-the-art methods with video-level features, and deliver competitive results with 22x faster inference time comparing with frame-level features.