A convolutional neural network program that identifies inverted Chinese character captchas used by Zhihu for login verification.
Zheye is a machine learning program specifically designed to recognize and locate inverted Chinese characters in the captcha images used by Zhihu for user login verification. It solves the problem of automated captcha decoding by applying convolutional neural networks and Gaussian mixture models to analyze and interpret these challenging visual puzzles.
Developers, data scientists, and machine learning enthusiasts interested in captcha recognition, computer vision, or experimenting with neural networks on real-world Chinese text datasets.
It provides a focused, open-source solution for a niche but practical captcha-breaking task, using modern deep learning frameworks like PyTorch, and serves as an educational resource for understanding image-based text recognition.
者也 - 知乎 倒立的文字 汉字验证码识别程序
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Accurately identifies and locates inverted Chinese characters in Zhihu captchas using a convolutional neural network, as highlighted in the key features for precise recognition.
Demonstrates practical use of CNN and Gaussian mixture models with PyTorch, providing hands-on learning for captcha recognition and image segmentation techniques.
Includes scripts for training and evaluation on real captcha samples, allowing users to replicate and experiment with the model, as shown in the usage instructions.
Explicitly states the project is for learning and entertainment, warning against misuse and encouraging compliance with terms of service, per the README's license section.
Designed solely for Zhihu's specific inverted Chinese character captchas, making it ineffective for general captcha solving without significant retraining and modification.
The README provides only basic usage commands and lacks detailed explanations, API references, or configuration guides, which limits deeper understanding and customization.
Targets a captcha format that Zhihu might change, and the struck-through online evaluation link indicates limited maintenance, reducing long-term reliability.