Diffusion Models as Masked Audio-Video Learners

TL;DR

This paper explores integrating diffusion models into the MAViL framework to improve efficiency in audio-visual pre-training, achieving significant reductions in computational cost without sacrificing downstream task performance.

Contribution

It introduces a novel combination of diffusion models with MAViL, enhancing training efficiency through curriculum masking and adaptive batch sizing.

Findings

32% reduction in pre-training FLOPS

18% decrease in pre-training wall clock time

Maintains performance on downstream audio classification

Abstract

Over the past several years, the synchronization between audio and visual signals has been leveraged to learn richer audio-visual representations. Aided by the large availability of unlabeled videos, many unsupervised training frameworks have demonstrated impressive results in various downstream audio and video tasks. Recently, Masked Audio-Video Learners (MAViL) has emerged as a state-of-the-art audio-video pre-training framework. MAViL couples contrastive learning with masked autoencoding to jointly reconstruct audio spectrograms and video frames by fusing information from both modalities. In this paper, we study the potential synergy between diffusion models and MAViL, seeking to derive mutual benefits from these two frameworks. The incorporation of diffusion into MAViL, combined with various training efficiency methodologies that include the utilization of a masking ratio curriculum and adaptive batch sizing, results in a notable 32% reduction in pre-training Floating-Point Operations (FLOPS) and an 18% decrease in pre-training wall clock time. Crucially, this enhanced efficiency does not compromise the model's performance in downstream audio-classification tasks when compared to MAViL's performance.