Optical character Detection & Recognition

Introduction

OCR is the method to detect text from images or live cam and convert it to machine-encoded text.I have used OpenCV's EAST(Efficient and Accurate Scene Text Detection), which is a deep learning model created to obtain high accuracy in text detection

Requirements

You need:

• OpenCV
• Python
• numpy
• EAST model: Link to download

Run the code

• Download and place the .pb file as the same folder as this
• The options provided are:

usage: ocr.py [-h] [-i IMAGE] [-east EAST] [-c MIN_CONFIDENCE] [-w WIDTH] [-e HEIGHT]

optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
  -east EAST, --east EAST
  -c MIN_CONFIDENCE, --min-confidence MIN_CONFIDENCE
  -w WIDTH, --width WIDTH
  -e HEIGHT, --height HEIGHT

Limitations

As of now, it does detect text but not completely.As seen in the below examples

To-do

Implement character-recognition from the detected text,using Tessaract or other OCR engine

Reference

Pyimagesearch Blog