kerlomz/captcha_trainer
Deep learning captcha and OCR training system using CNN/ResNet+GRU/LSTM+CTC architecture on TensorFlow.
Not currently ranked — collecting fresh signals.
star history
This project provides a deep learning solution for captcha recognition using TensorFlow, supporting complex scenarios like character粘连, perspective distortion, blur, and noise. It uses a three-layer architecture: convolutional layers for feature extraction from input images, recurrent layers (GRU/LSTM) for predicting label distributions from feature sequences, and a transcription layer using CTC for decoding final results by deduplication and integration.
Frequently asked
- What is kerlomz/captcha_trainer?
- Deep learning captcha and OCR training system using CNN/ResNet+GRU/LSTM+CTC architecture on TensorFlow.
- Is captcha_trainer open source?
- Yes — kerlomz/captcha_trainer is open source, released under the Apache-2.0 license.
- What language is captcha_trainer written in?
- kerlomz/captcha_trainer is primarily written in Python.
- How popular is captcha_trainer?
- kerlomz/captcha_trainer has 3.2k stars on GitHub.
- Where can I find captcha_trainer?
- kerlomz/captcha_trainer is on GitHub at https://github.com/kerlomz/captcha_trainer.