OCR on Hand Written Characters
Category: Success Stories

Success Stories

Client Overview

A Japanese OCR company.

Business Challenge

Client wanted to develop OCR which can recognize handwritten text with high accuracy. As handwritten text are not standard and uniform, client wanted to have deep learning framework for OCR to recognize characters.

Acclivis Contribution

  • CNN Layers library design and implementation.
  • Dataset Collection and generation.
  • Design and Implementation of CNN Net.
  • Training and Testing of CNN model.
  • CNN trained model generation.

Product Features Developed/Supported by Acclivis

  • The application performs handwritten digit and English character recognition using
    deep learning CNN network and model in Matlab.
  • CNN model is generated/trained with given NET architecture using 60K handwritten
    digits and English characters.

Project Details

  • Language | Matlab.
  • Platform | Intel CORE I3 without GPU.
  • DL Network Layers | Layer Network.
  • Classifier | SoftMax.

Client Benefit

Acclivis successfully delivered deep learning based OCR which could be integrated in client’s existing framework.

System Architecture

Hand Sign Detection
Previous
Translate »