----------------------------------------------
Lexicons for Cropped Word Image Datasets (Roy et al., ACCV 14)

Website: http://cvit.iiit.ac.in/projects/SceneTextUnderstanding/
----------------------------------------------

Datasets included: ICDAR 03, ICDAR 11, ICDAR 13, IIIT5K-word, SVT
Lexicon types and sizes: Small (50), Medium (1000), Large (0.5M)

Notes:
- ICDAR 03 small lexicon as shared by [2]
- IIIT 5K small/medium lexicon as shared in the dataset
- Large lexicon[4] common to all datasets 
- Lexicons generated using words semantically related to the ground truth from the WordNet API [3]
- Please see our ACCV 14 poster for results on the various datasets with the lexicons
- If using the lexicons, please cite our ACCV 14 paper,

@inproceedings{royscene,
  title={Scene Text Recognition and Retrieval for Large Lexicons},
  author={Roy, Udit and Mishra, Anand and Alahari, Karteek and Jawahar, CV},
  booktitle={ACCV-Asian Conference on Computer Vision}
}


Usage:
----------------------------------------------

- large_lexicon.mat - cell array of 0.5M words including ground truths from various datasets (common to all datasets)
- <dataset_name>.mat - contains a structure array with the following fields,
>> load ICDAR11.mat
>> testdata(1)

ans = 

        ImgName: 'test/268_0.jpg' 		% test image name
    GroundTruth: '2002' 			% ground truth label
      smallLexi: {1x50 cell} 			% cell array of 50 words
     mediumLexi: {1x1000 cell} 			% cell array of 1000 words


Contact:
----------------------------------------------
For any queries about the dataset contact: {udit.roy, anand.mishra}@research.iiit.ac.in


References:
----------------------------------------------
[1] Roy, U., Mishra, A., Alahari, K., & Jawahar, C. V. Scene Text Recognition and Retrieval for Large Lexicons. In ACCV-Asian Conference on Computer Vision. 

[2] Wang, K., Belongie, S.: Word Spotting in the Wild. In: ECCV. (2010)

[3] Miller, George A. "WordNet: a lexical database for English." Communications of the ACM 38.11 (1995): 39-41.

[4] Weinman, J. J., Learned-Miller, E., & Hanson, A. R. (2009). Scene text recognition using similarity and a lexicon with sparse belief propagation. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 31(10), 1733-1746.
