Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition
About
In this work we present a framework for the recognition of natural scene text. Our framework does not require any human-labelled data, and performs word recognition on the whole image holistically, departing from the character based recognition systems of the past. The deep neural network models at the centre of this framework are trained solely on data produced by a synthetic text generation engine -- synthetic data that is highly realistic and sufficient to replace real data, giving us infinite amounts of training data. This excess of data exposes new possibilities for word recognition models, and here we consider three models, each one "reading" words in a different way: via 90k-way dictionary encoding, character sequence encoding, and bag-of-N-grams encoding. In the scenarios of language based and completely unconstrained text recognition we greatly improve upon state-of-the-art performance on standard datasets, using our fast, simple machinery and requiring zero data-acquisition costs.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Scene Text Recognition | SVT (test) | Word Accuracy95.4 | 289 | |
| Scene Text Recognition | IIIT5K (test) | Word Accuracy97.1 | 244 | |
| Scene Text Recognition | SVT 647 (test) | Accuracy68 | 101 | |
| Text Detection | ICDAR 2013 (test) | F1 Score84.1 | 88 | |
| Scene Text Recognition | IC03 (test) | Accuracy98.7 | 63 | |
| Scene Text Recognition | IC13 1015 (test) | Accuracy79.5 | 27 | |
| Scene Text Recognition | ICDAR case-insensitive 2013 (test) | Accuracy90.8 | 22 | |
| Scene Text Recognition | ICDAR 50-word lexicon 2003 (test) | Accuracy99.2 | 16 | |
| Scene Text Recognition | Street View Text 50-word lexicon (test) | Accuracy0.961 | 15 | |
| Scene Text Recognition | ICDAR 2003 Full lexicon (test) | Accuracy98.6 | 11 |