Sequential Contrastive Audio-Visual Learning
Ioannis Tsiamas, Santiago Pascual, Chunghsin Yeh, Joan Serrà
TL;DR
The paper tackles the loss of fine-grained temporal information in conventional audio-visual contrastive learning by moving from aggregated embeddings to non-aggregated sequential representations. It introduces sequential contrastive learning (SCAV), which uses a distance-based objective on sequence latent spaces, with distances such as the Interpolated Euclidean, DTW, and Wasserstein, and applies row- and column-wise normalization to emphasize correct pairings. Empirically, SCAV with interpolated Euclidean distance achieves up to 2–3.5x recall gains on VGGSound and Music compared to aggregation-based methods while using fewer training resources, and demonstrates robust retrieval performance across metrics via a hybrid retrieval strategy. The work suggests that explicit intra-sequence modeling yields stronger multimodal representations and offers scalable retrieval, with potential extensions to incorporate text modalities for broader multimodal applications.
Abstract
Contrastive learning has emerged as a powerful technique in audio-visual representation learning, leveraging the natural co-occurrence of audio and visual modalities in webscale video datasets. However, conventional contrastive audio-visual learning (CAV) methodologies often rely on aggregated representations derived through temporal aggregation, neglecting the intrinsic sequential nature of the data. This oversight raises concerns regarding the ability of standard approaches to capture and utilize fine-grained information within sequences. In response to this limitation, we propose sequential contrastive audiovisual learning (SCAV), which contrasts examples based on their non-aggregated representation space using multidimensional sequential distances. Audio-visual retrieval experiments with the VGGSound and Music datasets demonstrate the effectiveness of SCAV, with up to 3.5x relative improvements in recall against traditional aggregation-based contrastive learning and other previously proposed methods, which utilize more parameters and data. We also show that models trained with SCAV exhibit a significant degree of flexibility regarding the metric employed for retrieval, allowing us to use a hybrid retrieval approach that is both effective and efficient.