cos120/captcha_crack
Free选字验证码破解，试验过网易和极验，破解率99
FreeFree tier
Inputs: image
About cos120/captcha_crack
An open-source tool for cracking click-select CAPTCHAs (such as NetEase and GeeTest) using YOLO for text localization and a trained classifier for recognition. Achieves a reported 99% success rate. Provides training scripts via run.sh and a demo server (python/server.py) for NetEase CAPTCHAs. Includes dataset references and tutorials. Licensed under multiple licenses; intended for academic use only.
Key Features
Uses YOLO for text localization in click-select CAPTCHAs
Trains a classifier for character recognition after localization
Includes training and recognition commands via run.sh
Provides a demo server (python/server.py) for NetEase click-select CAPTCHAs
Supports click-select CAPTCHAs from NetEase and GeeTest
Open-source with multiple license options (MIT, GPL-3.0, etc.)
Pros & Cons
Pros
- High reported success rate of 99% on tested platforms
- Open-source and free to use for academic purposes
- Combines localization (YOLO) and classification in one pipeline
- Includes a ready-to-use demo server for NetEase CAPTCHAs
- Provides detailed training and dataset instructions (run.sh, tutorials)
Cons
- Limited to click-select CAPTCHAs; not applicable to text-based or image recognition CAPTCHAs
- May require substantial training data and computational resources (GPU recommended)
- Prohibited for commercial use per disclaimer
- Documentation and comments primarily in Chinese, limiting accessibility
- Success rate may vary with different CAPTCHA implementations or after updates
Best For
Cracking click-select CAPTCHAs for security researchAcademic study of CAPTCHA robustness and vulnerabilityAutomated testing of CAPTCHA systemsDemonstrating YOLO-based object detection for text in images
FAQ
What types of CAPTCHAs does this tool crack?
It cracks click-select CAPTCHAs, specifically tested on NetEase and GeeTest (极验).
What is the claimed success rate?
The tool claims a 99% success rate based on testing.
Is the tool available for commercial use?
No, the project states it is for academic exchange only and must not be used for commercial purposes.
How do I use this tool?
Refer to run.sh for training and recognition commands. For a demo with NetEase CAPTCHAs, run python/server.py, which provides a URL-based recognition service.
What licenses are used?
The repository includes multiple license files (MIT, GPL-3.0, WTFPL, etc.). The applicable license may vary; check individual files.