TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models

TL;DR

TrOCR introduces an end-to-end Transformer-based approach for optical character recognition that leverages pre-trained models for improved accuracy across printed, handwritten, and scene text recognition tasks.

Contribution

The paper presents a novel Transformer-based OCR model that simplifies the pipeline by combining image understanding and text generation in a single end-to-end system.

Findings

Outperforms state-of-the-art models on multiple text recognition tasks

Effective use of pre-training with synthetic data improves accuracy

Applicable to printed, handwritten, and scene text recognition

Abstract

Text recognition is a long-standing research problem for document digitalization. Existing approaches are usually built based on CNN for image understanding and RNN for char-level text generation. In addition, another language model is usually needed to improve the overall accuracy as a post-processing step. In this paper, we propose an end-to-end text recognition approach with pre-trained image Transformer and text Transformer models, namely TrOCR, which leverages the Transformer architecture for both image understanding and wordpiece-level text generation. The TrOCR model is simple but effective, and can be pre-trained with large-scale synthetic data and fine-tuned with human-labeled datasets. Experiments show that the TrOCR model outperforms the current state-of-the-art models on the printed, handwritten and scene text recognition tasks. The TrOCR models and code are publicly available at \url{https://aka.ms/trocr}.