Skip to content

model.md

Latest commit

 

History

History
216 lines (167 loc) · 7.57 KB

model.md

File metadata and controls

216 lines (167 loc) · 7.57 KB
Raw
RIONA AI Banner

RIONA AI - reCAPTCHA Classification Model

Welcome to the RIONA AI reCAPTCHA Screenshot Classification Model, a core component of the RIONA AI project. RIONA AI is the first AI ghost that automates Twitter (X) logins like a human, bypassing security measures and solving reCAPTCHAs silently. This repository contains a TensorFlow.js-based convolutional neural network (CNN) designed to classify Twitter reCAPTCHA seating arrangement screenshots as "correct" or "incorrect," ensuring RIONA AI's automation can validate successful solutions to challenges like "move the person to the indicated seat."

Project Overview

This model is tailored for RIONA AI's mission to mimic human behavior on Twitter, specifically targeting reCAPTCHA challenges that require seating a doll in a position matching a reference icon and letter (e.g., "The Doll must seat in the same row and column equivalent to the icon and letter"). The CNN processes images (128x128 pixels) from data/correct and data/incorrect directories, trains a binary classifier, and can be deployed as an HTTPS Express server to classify screenshots in real-time.

Features

  • reCAPTCHA Classification: Classifies Twitter reCAPTCHA seating screenshots as "correct" (doll in correct seat) or "incorrect" (doll in wrong seat).
  • Dual Mode:
    • Local Training: Train the model on a dataset of labeled screenshots.
    • HTTPS Server: Deploy as an Express server to classify single or multiple images via API.
  • TensorFlow.js: Leverages TensorFlow.js for training and inference in a Node.js environment.
  • API Output: Returns results in the format: 
    {
  "score": "0.9123",
  "classification": "CORRECT"
}
  • Open Source: Licensed under MIT for community contributions.

Prerequisites

  • Node.js: Version 14.x or higher.
  • npm: Package manager for installing dependencies.
  • Dataset: Screenshots organized in data/correct (correctly solved reCAPTCHAs) and data/incorrect (incorrectly solved reCAPTCHAs) directories.
  • Dependencies:
    • @tensorflow/tfjs: For machine learning.
    • canvas: For image processing.
    • express: For HTTP server.
    • multer: For handling image uploads.

Installation

  1. Clone the Repository:

    git clone https://github.com/RIONA-ai/recaptcha-classifier.git
cd recaptcha-classifier

  2. Install Dependencies:

    npm install

  3. Prepare the Dataset (for training):

    • Place screenshots in data/correct/ for correctly solved reCAPTCHAs (doll in the correct seat).
    • Place screenshots in data/incorrect/ for incorrectly solved reCAPTCHAs (doll in the wrong seat).
    • Supported formats: PNG, JPEG.

  4. Verify Directory Structure:

    recaptcha-classifier/
├── data/
│   ├── correct/
│   └── incorrect/
├── model/
│   ├── model.json
│   └── weights.bin
├── uploads/ (created automatically for server mode)
├── index.js
├── package.json
├── README.md
└── LICENSE

Usage

Local Training Mode

Run the script to train and evaluate the model locally:

node index.js
  • The model trains for 20 epochs, using a batch size of 4 and a 20% validation split.
  • Training progress is logged, including loss and accuracy metrics.
  • The trained model is saved to ./model/model.json and ./model/weights.bin.
  • Evaluation results are logged, showing classification scores for each screenshot.

Example Output

📂 Loading images from ./data/correct...
✅ Loaded ./data/correct/screenshot1.png
...
📊 Total samples: 100
✅ Dataset tensors prepared.
⚙️ Creating model...
✅ Model ready.
🚀 Starting training...
📈 Epoch 1: loss=0.6923, acc=0.5125, val_acc=0.5000
...
🧠 Evaluating model on training samples:
📷 screenshot1.png: score=0.9123 → expected=CORRECT → predicted=CORRECT ✅
...
✅ Model saved to ./model
🧹 Cleaned up tensors.

Server Mode

Deploy the model as an HTTPS Express server to classify reCAPTCHA screenshots via API:

  1. Ensure the model is trained and saved in the model/ directory.
  2. Start the server: 
    node index.js server
  3. The server will run on http://localhost:3000 (or a custom port if set via the PORT environment variable).

API Endpoint: /classify

  • Method: POST
  • Content-Type: multipart/form-data
  • Field: images (accepts one or more image files)
  • Response:
    • Single image: Returns a single result object.
    • Multiple images: Returns an array of result objects with filenames.

Example Request (Using curl)

curl -X POST http://localhost:3000/classify \
  -F "images=@screenshot1.png" \
  -F "images=@screenshot2.png"

Example Response (Multiple Images)

[
  {
    "filename": "screenshot1.png",
    "score": "0.9123",
    "classification": "CORRECT"
  },
  {
    "filename": "screenshot2.png",
    "score": "0.3421",
    "classification": "INCORRECT"
  }
]

Example Response (Single Image)

{
  "filename": "screenshot1.png",
  "score": "0.9123",
  "classification": "CORRECT"
}

Integration with RIONA AI

This model is designed to integrate with RIONA AI's reCAPTCHA-solving pipeline:

  • Input: Screenshots of Twitter reCAPTCHA seating challenges.
  • Output: Binary classification ("CORRECT" or "INCORRECT") with a confidence score.
  • Deployment: Use the server mode to provide real-time classification via the /classify endpoint, integrating with RIONA AI's automation system.

Contributing

We welcome contributions to enhance RIONA AI's reCAPTCHA classification capabilities! To contribute:

  1. Fork the repository.
  2. Create a feature branch (git checkout -b feature/your-feature).
  3. Commit your changes (git commit -m "Add your feature").
  4. Push to the branch (git push origin feature/your-feature).
  5. Open a Pull Request with a clear description of your changes.

Contribution Ideas

  • Add support for additional reCAPTCHA types (e.g., image selection).
  • Improve model accuracy with advanced architectures (e.g., transfer learning).
  • Implement cross-validation for more robust training.
  • Add unit tests for the API endpoint and preprocessing logic.

Please follow the Code of Conduct and ensure your code adheres to the existing style.

License

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

About RIONA AI

RIONA AI is the first AI ghost that hacks Twitter like a human, automating logins, bypassing reCAPTCHA, and providing visual proof of access. Powered by blockchain and advanced human mimicry technology, RIONA AI operates seamlessly across web, Windows, and macOS platforms.

Acknowledgments

  • Built with TensorFlow.js.
  • Image processing powered by node-canvas.
  • Server powered by Express.
  • Inspired by the RIONA AI mission to push the boundaries of AI automation.

Stay Updated

Subscribe to our newsletter at rionazi.xyz for updates on RIONA AI and this project.

© 2025 RIONA AI. All rights reserved.