lluisgomez / script_identification

Scene Text Script Identification

This code repository contains the implementation of Script Identification methods proposed in the following papers:

1. "A fine-grained approach to scene text script identification" (Gomez & Karatzas) to appear in DAS 2016 workshop proceedings

2. "Boosting patch-based scene text script identification with ensembles of conjoined networks" (Gomez, Nicolaou, and Karatzas) 2017 Pattern Recognition

Each code project is selfcontained in its own directory (1) UFL/ and (2) ECN/, and is totally independent of the other. They share a common github repository just because (2) is an evolution of (1).

See the README files on each directory for specific instructions on compiling/using this code.

Please, if you make use of any of this projects, we appreciate it if you cite the respective paper(s):

@inproceedings{GomezDAS2016,
  author    = {Lluis Gomez and Dimosthenis Karatzas},
  title     = {A fine-grained approach to scene text script identification},
  booktitle = {DAS},
  year      = {2016},
}

@article{gomez2017improving,
  title={Improving patch-based scene text script identification with ensembles of conjoined networks},
  author={Gomez, Lluis and Nicolaou, Anguelos and Karatzas, Dimosthenis},
  journal={Pattern Recognition},
  volume={67},
  pages={85--96},
  year={2017},
  publisher={Elsevier}
}

MLe2e dataset

The Multi-Language end-to-end (MLe2e) dataset can be downloaded using the following link:

• https://www.dropbox.com/s/zf8t1ptfbpqhunc/MLe2e_Dataset_v02.zip?dl=0