ASAYAR: A Dataset for Arabic-Latin Scene Text Localization in Highway Traffic Panels



Overview

Welcome to ASAYAR, the first public dataset dedicated for Latin (French) and Arabic Scene Text Detection in Highway panels. It comprises more than 1800 well-annotated images. The dataset was colleted from Moroccan Highway and it has been manually annotated. ASAYAR data can be used to develop and evaluate traffic signs detection and French or Arabic text detection in different languages.

Examples of Annotated Images

Latin/Arabic Scene Text Detection Dataset


Citation

Our paper introducing the dataset and the evaluations methods is published at the IEEE Transactions on Intelligent Transportation Systems 2020 and available here. If you make use of the ASAYAR dataset, please cite our following paper: 

  @ARTICLE{9233923,
          author={M. {Akallouch} and K. S. {Boujemaa} and A. {Bouhoute} and K. {Fardousse} and I. {Berrada}},
          journal={IEEE Transactions on Intelligent Transportation Systems}, 
          title={ASAYAR: A Dataset for Arabic-Latin Scene Text Localization in Highway Traffic Panels}, 
          year={2020},
          pages={1-11},
          doi={10.1109/TITS.2020.3029451}}

Image Source and Usage License


The images in ASAYAR dataset are manily filmed from the Moroccan Highway.
All images and their associated annotations in ASAYAR can be used for academic purposes only, but any commercial use is prohibited. (read more)

Annotation format

In the dataset, each instance's location is annotated by a rectangle bounding boxes. The bounding box can be denoted as {XMIN, YMIN, XMAX, YMAX}. An obejct has a class name denoted as CLASS. The global image information are defined as follows: FOLDER, PATH, NAME, and SIZE.
The following script shows an example of annotation: 

  
  <annotation>
      <folder>FOLDER</folder>
      <filename>IMAGE_NAME</filename>
      <path>PATH</path>
      <source>
          <database>ASAYAR</database>
      </source>
      <size>
          <width>WIDTH</width>
          <height>HEIGHT</height>
          <depth>DEPTH</depth>
      </size>
      <object>
          <name>CLASS</name>
          <bndbox>
              <xmin>XMIN</xmin>
              <ymin>YMIN</ymin>
              <xmax>XMAX</xmax>
              <ymax>YMAX</ymax>
          </bndbox>
      </object>
      ...
  </annotation>

Dataset structure

  
  Train or Test/
  ├── Traffic signs/
  │   ├── Annotations/
  │   │   ├── image_1.xml
  │   │   └── ...
  │   └── Images
  │       ├── image_1.png
  │       └── ...
  │       
  ├── Text/
  │   ├── Word Level/
  │   │   ├── Annotations/
  │   │   │   ├── image_1.xml
  │   │   │   └── ...
  │   │   └── Images/
  │   │       ├── image_1.png
  │   │       └── ...
  │   └── Line Level/
  │       ├── Annotations/
  │       │   ├── image_1.xml
  │       │   └── ...
  │       └── Images/
  │           ├── image_1.png
  │           └── ...
  └── Directional Symbols/
      ├── Annotations/
      │   ├── image_1.xml
      │   └── ...
      └── Images/
          ├── image_1.png
          └── ...

Contact

For any questions, comments, suggestions or feedback, do not hesitate to send us an e-mail :

Sidi Mohamed Ben Abdellah University - Fez Morocco