Assignment 5. Object detection and document recognition

Overview

In this task, we have been required to recognize the content of a small text document. A receipt from supermarket or electricity bill cheque should be fine.

Setup

Prerequisites:

• Python 3.10 or higher
• Libraries: OpenCV, NumPy, Torch, torchvision, EasyOCR, Matplotlib
• A camera or an image file of a skewed document (e.g., receipt or bill)

Installation:

1. Clone the repository:

   git clone https://github.com/ADA-GWU/a5-object-detection-aliasgerovs.git
   cd a5-object-detection-aliasgerovs

2. Install the required Python packages:

   pip3 install -r requirements.txt

Running the Code

To run the text recognition:

Execute the script from the command line

   python3 recognition.py

3. Follow the on-screen instructions to select the corners of the document in the displayed window.
4. Press "p" to process the image after selecting the corners.

Features

• Perspective Transformation: Converts the trapezoidal shape of the document to a rectangular form using manually selected corners.
• Text Detection: Utilizes the MSER algorithm to detect regions likely to contain text.
• Text Recognition: Applies a pre-trained CNN to recognize the detected text.
• Visualization: Shows the processed images and detected text regions with recognized text.

Code Structure

• text_recognition.py: Contains the main logic for image processing, text detection, and recognition.
• images folder: Contains images.