https://github.com/tautcony/captcha-tfjs-demo

Run an OCR model in the browser via tfjs
https://github.com/tautcony/captcha-tfjs-demo

captcha tensorflow tensorflowjs

Last synced: about 1 month ago
JSON representation

Run an OCR model in the browser via tfjs

• Host: GitHub
• URL: https://github.com/tautcony/captcha-tfjs-demo
• Owner: tautcony
• License: gpl-3.0
• Created: 2024-07-19T13:56:22.000Z (6 months ago)
• Default Branch: master
• Last Pushed: 2024-07-24T08:04:27.000Z (6 months ago)
• Last Synced: 2024-07-24T09:28:25.044Z (5 months ago)
• Topics: captcha, tensorflow, tensorflowjs
• Language: Python
• Homepage:
• Size: 2.9 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# TensorFlow.js Captcha Recognition

This project aims to train and use TensorFlow.js to perform captcha recognition in the browser.

## Table of Contents

- [Introduction](#introduction)

- [Usage](#usage)

- [License](#license)

## Introduction

In this project, we leverage the power of TensorFlow to train a machine learning model for captcha recognition. By using TensorFlow.js, we can perform captcha recognition directly in the browser without the need for server-side processing.

## Usage

Note: In this project, all captcha images are located in the `data` folder and are labeled by their filenames, with the following deformation characteristics:

1. The characters are surrounded by a lot of noise

1. The individual characters themselves are slightly distorted.

![captcha example](./test/2589.png)

To implement the captcha recognition functionality in your browser, follow these steps:

1. prepare and label the captcha images for training

1. run `0-preprocess.py` to preprocess the captcha images

1. run `1-generate_dataset.py` to generate the dataset

1. run `2-train_model.py` to train the model

1. run `3-test_model.py` to test the model

1. run `4-convert_tfjs_model.py` to convert model for tensorflow.js

Now you have the model in the `web/tfjs_model` folder, start a server under the `web` directory, you can now perform captcha recognition in the browser.

## License

Distributed under the GPLv3+ License. See LICENSE for more information.