Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data

TL;DR

This paper introduces a generative shape model that excels in scene text recognition and segmentation, achieving high accuracy with significantly less training data and demonstrating robustness to various transformations.

Contribution

The paper presents a novel generative shape model that performs joint text recognition and segmentation with minimal training data, outperforming existing discriminative methods.

Findings

Achieves state-of-the-art scene text recognition results

Requires orders of magnitude fewer training images

More robust to affine and non-affine transformations

Abstract

We demonstrate that a generative model for object shapes can achieve state of the art results on challenging scene text recognition tasks, and with orders of magnitude fewer training images than required for competing discriminative methods. In addition to transcribing text from challenging images, our method performs fine-grained instance segmentation of characters. We show that our model is more robust to both affine transformations and non-affine deformations compared to previous approaches.