LoGAN: Latent Graph Co-Attention Network for Weakly-Supervised Video Moment Retrieval
This addresses the problem of localizing video segments from text queries without temporal annotations, offering a novel approach for video understanding applications.
The paper tackles weakly-supervised video moment retrieval by proposing LoGAN to model relational context between video frames using latent graph co-attention, achieving an 11% improvement in Recall@1 accuracy over strongly-supervised state-of-the-art methods on the DiDeMo dataset.
The goal of weakly-supervised video moment retrieval is to localize the video segment most relevant to the given natural language query without access to temporal annotations during training. Prior strongly- and weakly-supervised approaches often leverage co-attention mechanisms to learn visual-semantic representations for localization. However, while such approaches tend to focus on identifying relationships between elements of the video and language modalities, there is less emphasis on modeling relational context between video frames given the semantic context of the query. Consequently, the above-mentioned visual-semantic representations, built upon local frame features, do not contain much contextual information. To address this limitation, we propose a Latent Graph Co-Attention Network (LoGAN) that exploits fine-grained frame-by-word interactions to reason about correspondences between all possible pairs of frames, given the semantic context of the query. Comprehensive experiments across two datasets, DiDeMo and Charades-Sta, demonstrate the effectiveness of our proposed latent co-attention model where it outperforms current state-of-the-art (SOTA) weakly-supervised approaches by a significant margin. Notably, it even achieves a 11% improvement to Recall@1 accuracy over strongly-supervised SOTA methods on DiDeMo.