VideoBERT: A Joint Model for Video and Language Representation Learning

TL;DR

VideoBERT introduces a self-supervised, joint visual-linguistic model based on BERT to learn high-level video and language features, enabling improved performance in tasks like captioning and open-vocabulary classification.

Contribution

It is the first to adapt BERT for joint video and language representation learning using vector quantization and speech recognition data without explicit supervision.

Findings

Outperforms state-of-the-art in video captioning

Enables open-vocabulary classification directly from learned features

Shows that large data and cross-modal info improve performance

Abstract

Self-supervised learning has become increasingly important to leverage the abundance of unlabeled data available on platforms like YouTube. Whereas most existing approaches learn low-level representations, we propose a joint visual-linguistic model to learn high-level features without any explicit supervision. In particular, inspired by its recent success in language modeling, we build upon the BERT model to learn bidirectional joint distributions over sequences of visual and linguistic tokens, derived from vector quantization of video data and off-the-shelf speech recognition outputs, respectively. We use VideoBERT in numerous tasks, including action classification and video captioning. We show that it can be applied directly to open-vocabulary classification, and confirm that large amounts of training data and cross-modal information are critical to performance. Furthermore, we outperform the state-of-the-art on video captioning, and quantitative results verify that the model learns high-level semantic features.