No-Puzzle-Captcha

Lightweight Library for Slide Puzzle Captcha Solving
轻量级的滑动拼图验证码自动识别库

Introduction

This Python library is designed to solve puzzle CAPTCHAs (aka slide CAPTCHAs), using OpenCV to achieve high inference speed with low computational cost.

Features

• Fast : Average processing time is less than 5ms.
• Lightweight : CPU-friendly. No neural network.
• Minimal-Dependency : The only direct dependency is opencv-python.
• Well-Designed : Follows best practices such as type annotations.

Get Started

Installation

Install from PyPI:

pip install no_puzzle_captcha

Minimal Example

Firstly, prepare an background image (the original image) and a puzzle image (the slider image).

You only need 4 lines of code to solve a puzzle CAPTCHA:

from no_puzzle_captcha import PuzzleCaptchaSolver

solver = PuzzleCaptchaSolver()

result = solver.handle_file("background.png", "puzzle.png")

print(f"Matched at ({result.x}, {result.y})")

Usage

Solver Class

All the following methods in the class PuzzleCaptchaSolver returns a PuzzleCaptchaResult object:

• Method handle_file: Accepts two file path parameters (one for background image and another for puzzle image).
• Method handle_bytes: Accepts two bytes-like parameters that store the image data.
• Method handle_image: Accepts two matrix-like parameters that store the pixel data of the image.

Result Class

You can get the recognition result via the class PuzzleCaptchaResult:

• Property x and y: The top-left coordinates of the detected result.

You can also visualize the result.

• Method visualize: Returns a matrix-like object representing the visualized image.
• Method visualize_and_show: Shows the visualized image via OpenCV's window.
• Method visualize_and_save: Saves the visualized image to a specified path.

Advanced Usage

The constructor of the class PuzzleCaptchaSolver accepts an optional transforms argument, allowing you to customize the image transformation process.

Benchmark

The following test results are based on version 1.1.1. The elapsed time may vary depending on the device. You can run test.py to reproduce these tests.

GeeTest Test

This test uses the standard samples from GeeTest, each contains exactly one hollow in the background image.

Sample:

Background Image	Puzzle Image

Result:

Tricky Test

This test uses the samples with multiple hollow for misleading purpose, so we call it "tricky".

Sample:

Background Image	Puzzle Image

Result:

Tricky Hard Test

This test uses the samples with multiple hollow, but they are more difficult to recognize.

Sample:

Background Image	Puzzle Image

Result:

Credits

• OpenCV - provides a powerful computer vision tool.
• Vasyl Smutok's Puzzle-Captcha-Solver - gave me an inspiration.
• GeeTest | 2Captcha | GoCaptcha - provides samples of GeeTest-style puzzle CAPTCHAs.
• Capy - provides samples of difficult puzzle CAPTCHAS.
• USTB's API - provides samples of tricky puzzle CAPTCHAs.
• Chaoxing's API - provides samples of tricky-hard puzzle CAPTCHAs.

Licensing

This project is licensed under the MIT License. See the License file for more details.