Cascaded Multilingual Audio-Visual Learning from Videos

TL;DR

This paper introduces a cascaded multilingual audio-visual learning approach that leverages English-trained models to improve cross-lingual retrieval and captioning performance on videos in other languages, achieving significant gains.

Contribution

The paper presents a novel cascaded method for multilingual audio-visual learning that enhances cross-lingual transfer from English-trained models to other languages.

Findings

Nearly 10x improvement in retrieval performance on Japanese videos

State-of-the-art results on Japanese and Hindi spoken caption tasks

Effective cross-lingual transfer from English to other languages

Abstract

In this paper, we explore self-supervised audio-visual models that learn from instructional videos. Prior work has shown that these models can relate spoken words and sounds to visual content after training on a large-scale dataset of videos, but they were only trained and evaluated on videos in English. To learn multilingual audio-visual representations, we propose a cascaded approach that leverages a model trained on English videos and applies it to audio-visual data in other languages, such as Japanese videos. With our cascaded approach, we show an improvement in retrieval performance of nearly 10x compared to training on the Japanese videos solely. We also apply the model trained on English videos to Japanese and Hindi spoken captions of images, achieving state-of-the-art performance.