Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding

TL;DR

Pix2Struct is a versatile pretrained model that parses masked web page screenshots into HTML, enabling it to excel across diverse visually-situated language tasks by unifying multiple signals like OCR and captioning.

Contribution

The paper introduces Pix2Struct, a novel pretraining approach using screenshot parsing into HTML, with flexible input integration, achieving state-of-the-art results across multiple domains.

Findings

Achieves state-of-the-art on six out of nine tasks.

Effective across documents, illustrations, UIs, and natural images.

Demonstrates the power of visual language pretraining with HTML parsing.

Abstract

Visually-situated language is ubiquitous -- sources range from textbooks with diagrams to web pages with images and tables, to mobile apps with buttons and forms. Perhaps due to this diversity, previous work has typically relied on domain-specific recipes with limited sharing of the underlying data, model architectures, and objectives. We present Pix2Struct, a pretrained image-to-text model for purely visual language understanding, which can be finetuned on tasks containing visually-situated language. Pix2Struct is pretrained by learning to parse masked screenshots of web pages into simplified HTML. The web, with its richness of visual elements cleanly reflected in the HTML structure, provides a large source of pretraining data well suited to the diversity of downstream tasks. Intuitively, this objective subsumes common pretraining signals such as OCR, language modeling, image captioning. In addition to the novel pretraining strategy, we introduce a variable-resolution input representation and a more flexible integration of language and vision inputs, where language prompts such as questions are rendered directly on top of the input image. For the first time, we show that a single pretrained model can achieve state-of-the-art results in six out of nine tasks across four domains: documents, illustrations, user interfaces, and natural images.