The SVT-CHAR dataset contains character level bounding boxes 
and ground truth annotation of SVT-WORD dataset.

How to load character Bounding Box information
-----------------------------------------------------------
(Usage: Case sensitive character detection/recognition)

1. Open Matlab
2. load testCharBound 
3. A structure testCharBound will be loaded. 
This structure has three fields
   (i) ImgName: Specifies SVT-WORD cropped word image name. 
   (ii) chars: Contained a string of characters.
   (iii) charBB: Bounding box of characters in same order 
         as chars. Bounding box have format [x y width height].
-------------------------------------------------------------  

Example:
>> load testCharBound

>> testCharBound(1).ImgName

ans =

00_02_1.png

>> testCharBound(1).chars

ans =

FORUM

>> testCharBound(1).charBB %% Loads bounding box for characters F, O, R, U, M in [x y width height] format

ans =

     9     9    33    63
    44    10    43    64
    91    10    40    63
   133    10    37    64
   175    11    45    64


-------------------------------------------------------------------
If you use this dataset, please cite:

@InProceedings{Mishra12,
  author    = "Mishra, A. and Alahari, K. and Jawahar, C.~V.",
  title     = "Top-Down and Bottom-Up Cues for Scene Text Recognition",
  booktitle = "CVPR",
  year      = "2012"
}


Since this dataset is originally taken from: http://vision.ucsd.edu/~kai/svt/. 
The following paper should be cited when using this dataset.

@InProceedings{Wang11,
  author    = "Wang, K. and Babenko, B. and Belongie, S.",
  title     = "End-to-end Scene Text Recognition",
  booktitle = "ICCV",
  year      = "2011"
}
 

For any queries about the dataset contact: anand.mishra@research.iiit.ac.in