ocr for captcha

Introduction

This repository hosts a Keras implementation of an OCR model designed for reading CAPTCHAs. The model combines convolutional and recurrent neural networks to effectively translate CAPTCHA images into text.

Architecture

The model is constructed using the Keras Functional API, integrating CNN and RNN layers. It also demonstrates the use of a custom layer as an endpoint for computing CTC (Connectionist Temporal Classification) loss. The architecture leverages subclassing for creating custom layers, providing flexibility and modularity in model design.

Training

The training process involves configuring the model to recognize and decode CAPTCHA images. The example provided in the repository is a practical demonstration of how to train the model, using a combination of convolutional layers for feature extraction and recurrent layers for sequence prediction.

Guide: Running Locally

1. Clone the repository from the Hugging Face model card page.
2. Set up your environment with the necessary dependencies, preferably using a virtual environment.
3. Use the provided Jupyter notebook to run the example and experiment with the model.
4. For accelerated training and inference, consider using cloud GPU services such as Google Colab or AWS EC2 with GPU instances.

License

The model and its associated code are released under the CC0-1.0 license, allowing for unrestricted use, distribution, and modification.