Masked and Permuted Implicit Context Learning for Scene Text Recognition
This work addresses scene text recognition for applications like document analysis and autonomous systems, representing an incremental improvement over existing methods.
The paper tackled the problem of scene text recognition by addressing limitations in existing methods that use permuted or masked language modeling, proposing a unified approach that combines both within a single decoder. The result was superior performance on common benchmarks and a 9.1% improvement on the challenging Union14M-Benchmark.
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.