CardNumber-DeepOCR

Iranian Credit card and National card Deep Optical Character Recognition (OCR) using Keras.

Dataset is synthesized using OpenCV.

Table of Contents

1. Dataset
2. Model
3. How to run
4. Results
5. App Demo

Dataset

Iranian Credit Card (Different Banks)

Mellat Bank	Melli Bank

Synthesized credit cards images

Iranian National card

Old ones	New ones

Synthesized national cards images

Model

1. Synthesizing data and pre-processing [Notebook]
2. Classify the image (Credit or National) [Notebook]
3. Feed images to OCR network [Notebook]
4. Wrapping the pipeline [Notebook]

Classifier

OCR

How to run

For testing open E2E_Wrapper.ipynb inside Google Colab or open e2e_wrapper.py

API's:

Predict on a single image

predict_single_path(path, label=None)

Predict on a single image and show image, ground truth

single_file_show(path, label)

Predict on all images inside a folder path and report some metrics:

predict_folder(folder_path)

Results

Credit cards	National cards

App Demo

App_Demo.mp4