Next-generation character recognition using machine learning.
Overview
In this article, I am going to develop a React + FastAPI application that captures images from the webcam and uses an OCR-trained model to detect the text.
PaddleOCR is the OCR toolkits that I use. From my testing, it is a lightweight and high-performing ML OCR library. It supports over 80 languages and can be deployed in server, mobile, embedded, and IoT devices using languages like Python, C++, etc.
The Application
The application captures the image using the webcam and invokes the backend API to recognize the text.
Regions, where the characters are detected, are boxed, and the results are displayed with the probabilities.
More examples are shown below.
News and Document
Invoice
Number Plate
CAPTCHA
Custom Font
The Code
Backend
The backend APIs are developed using FastAPI. WebSocket is used to send and receive data.
