Arabic OCR

• OCR system for Arabic language that converts images of typed text to machine-encoded text.
  
• The system aims to solve a simpler problem of OCR with images that contain only Arabic characters (check the dataset link below to see a sample of the images).

Important Note

The system currently supports only letters (29 letters) ا-ى , لا (no numbers or special symbols).

Setup

Install python then run this command:

pip install -r requirements.txt

Run

1. Put the images in src/test directory
2. Go to src directory and run the following command

   python OCR.py
   

3. Output folder will be created with:
   • text folder which has text files corresponding to the images.
   • running_time file which has the time taken to process each image.

Pipeline

Dataset

• Link to dataset of images and the corresponding text: here.
• We used 1000 images to generate character dataset that we used for training.

Examples

Line Segmentation

Word Segmentation

Character Segmentation

Testing

NOTE: Make sure you have a folder with the truth output with same file names to compare it with the predicted text.

From within src folder run:

python edit.py 'output/text' 'truth'

Performance

• Average accuracy: 95%.
• Average time per image: 16 seconds.

NOTE

We achieved these results when we used only the flatten image as feature.

References

1. An Efficient, Font Independent Word and Character Segmentation Algorithm for Printed Arabic Text.

2. A Robust Line Segmentation Algorithm for Arabic Printed Text with Diacritics.

3. Arabic Character Segmentation Using Projection Based Approach with Profile's Amplitude Filter .