OCR with image processing and deep learning
Last updated on 2020.09.24.
Tüzes Izabella
Other, HR, Logistics, Agriculture, Industry, Bank Insurance, IT, Telecom, Solutions
OCR provides us with different ways to see an image
Almost all companies -- especially certain ones in some service areas -- struggle with processing inundating volume of incoming paper documents. In most cases this requires time consuming and costly human work of various quality.
Example
The most typical task is extracting relevant pieces of information from digitalized papers and channel them to traditional tabular data tables or spreadsheets for storing or further processing; this is where optical character recognition comes into the picture with the help of image processing and deep learning. The actual implementation and the difficulty level depend on the nature of the specific task: whether the document contains handwritten or printed texts, we face with some kind of known predefined layouts or any kind of formats. The approach used to have different stage: first localization of the relevant texts (like filled out formulars) on the document at large, then the actual character recognition to turn pixels into strings, then optionally some post processing -- like dictionary or rule-based corrections -- for improving the final results.
Added values (Why AI/ML/DL): automatization of paper document processing for saving cost and time.
Proposed tech stack: Linux, Python (Anaconda), Scikit-learn, OpenCV, Pillow, TensorFlow, PyTorch, Tesseract
Search
Tech Stack
Android
Angular
Apache
Artificial Intelligence
Business Analysis
C++
Centura
Cloud
Docker
EAPware
Firebase
Gitlab
Industry 4.0
IoT
Java
JavaScript
JIRA
Linux
Load Test
M2M
MariaDB
Microservice
NLTK
Node.js
OpenWRT
Oracle
PHP
PLC programming
Python
PyTorch
RASA
Scikit-learn
TensorFlow
Test Automation