Masked and Permuted Implicit Context Learning for Scene Text Recognition

TL;DR

This paper introduces a novel scene text recognition model that combines permuted and masked language modeling techniques within a single decoder, improving robustness and accuracy on standard benchmarks.

Contribution

It unifies PLM and MLM in one decoder, incorporates word length info, and employs perturbation training for enhanced robustness in scene text recognition.

Findings

Achieves superior performance on standard benchmarks.

Improves 9.1% on Union14M-Benchmark.

Demonstrates robustness against length prediction errors.

Abstract

Scene Text Recognition (STR) is difficult because of the variations in text styles, shapes, and backgrounds. Though the integration of linguistic information enhances models' performance, existing methods based on either permuted language modeling (PLM) or masked language modeling (MLM) have their pitfalls. PLM's autoregressive decoding lacks foresight into subsequent characters, while MLM overlooks inter-character dependencies. Addressing these problems, we propose a masked and permuted implicit context learning network for STR, which unifies PLM and MLM within a single decoder, inheriting the advantages of both approaches. We utilize the training procedure of PLM, and to integrate MLM, we incorporate word length information into the decoding process and replace the undetermined characters with mask tokens. Besides, perturbation training is employed to train a more robust model against potential length prediction errors. Our empirical evaluations demonstrate the performance of our model. It not only achieves superior performance on the common benchmarks but also achieves a substantial improvement of $9.1\%$ on the more challenging Union14M-Benchmark.