https://github.com/hadim/captchanet

A simple but yet efficient neural networks to solve captcha images.
https://github.com/hadim/captchanet

Last synced: about 1 year ago
JSON representation

A simple but yet efficient neural networks to solve captcha images.

• Host: GitHub
• URL: https://github.com/hadim/captchanet
• Owner: hadim
• License: bsd-3-clause
• Created: 2019-07-01T18:19:12.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2019-07-26T13:04:16.000Z (almost 7 years ago)
• Last Synced: 2025-03-15T14:01:44.788Z (about 1 year ago)
• Language: Jupyter Notebook
• Size: 7.45 MB
• Stars: 1
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# captchanet

[![Build Status](https://travis-ci.com/hadim/captchanet.svg?branch=master)](https://travis-ci.com/hadim/captchanet)

A simple but yet efficient neural networks to solve captcha images.

Features:

- Generate your own dataset of random captcha images: no need to manually annotate a dataset, you only need to generate images as close as possible as the real ones.

- Use Tensorflow 2.0 and the Keras API.

- After training, model is exported as a Keras HDF5 model (soon it will be exported as a Saved Model file).

- Model can be loaded in an environment containing only Tensorflow. Useful for inference (no need to install the `captachanet` library).

![screenshot](screenshot.png "Screenshot")

## Install

Dependencies should be installed using Anaconda:

```bash

conda create -n captchanet_env

conda activate captchanet_env

conda env update -f environment.yml

# Then install libraries not available in conda-forge.

pip install --no-deps -U tensorflow-datasets tensorflow_metadata tensorboard tensorflow-estimator

pip install --no-deps -U tensorflow-gpu==2.0.0-beta1

pip install --no-deps -U https://github.com/keras-team/keras-tuner/archive/fe5ed591bf54fc3db033814ab43193f6e8999ee5.zip

```

Then install captchanet:

`pip install captchanet`

## Usage

As a starting point you can browser [the notebooks](./notebooks) and adapt it to our needs.

## License

Under BSD license. See [LICENSE](LICENSE).

## Authors

- Hadrien Mary