Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks

TL;DR

BEiT-3 is a versatile multimodal foundation model that leverages Multiway Transformers and unified masked modeling to achieve state-of-the-art results across a wide range of vision and vision-language tasks.

Contribution

The paper introduces BEiT-3, a general-purpose multimodal model with a modular architecture and unified pretraining task, advancing the convergence of vision and language modeling.

Findings

State-of-the-art performance on object detection and segmentation

Superior results in image classification and visual reasoning

Effective cross-modal retrieval and captioning

Abstract

A big convergence of language, vision, and multimodal pretraining is emerging. In this work, we introduce a general-purpose multimodal foundation model BEiT-3, which achieves state-of-the-art transfer performance on both vision and vision-language tasks. Specifically, we advance the big convergence from three aspects: backbone architecture, pretraining task, and model scaling up. We introduce Multiway Transformers for general-purpose modeling, where the modular architecture enables both deep fusion and modality-specific encoding. Based on the shared backbone, we perform masked "language" modeling on images (Imglish), texts (English), and image-text pairs ("parallel sentences") in a unified manner. Experimental results show that BEiT-3 obtains state-of-the-art performance on object detection (COCO), semantic segmentation (ADE20K), image classification (ImageNet), visual reasoning (NLVR2), visual question answering (VQAv2), image captioning (COCO), and cross-modal retrieval (Flickr30K, COCO).