/data-aug

Primary LanguageCSS

AugmentAI - Data Augmentation Tool

AugmentAI Cover

This project provides an image augmentation tool using Generative AI models, specifically designed and optimized for NVIDIA AI Workbench.

Table of Contents

Description

AugmentAI allows users to upload image datasets and augment them using generative models. The system uses Stability AI API for high-quality image generation and falls back to a local GAN model when needed. Built specifically for NVIDIA AI Workbench, the application utilizes GPU acceleration for fast and efficient image processing.

Demo

Watch the demonstration video here to see AugmentAI in action, including setup in NVIDIA AI Workbench and usage examples.

Features

  • Infrastructure

    • Built specifically for NVIDIA AI Workbench
    • Easy deployment and management
    • No complex setup needed

  • User Experience

    • Simple, clean interface
    • Folder upload support
    • Maintains original image dimensions
    • Direct result downloads

  • Future Potential

    • Framework ready for multi-modal expansion
    • Could integrate multiple AI models
    • Potential for text and tabular data support

Environment Setup

Required Environment Variables

  • STABILITY_API_KEY: Required for high-quality image generation (Default demo key is included for the judges)
    • Can be obtained from Stability AI platform
    • Set in Workbench environment or .env file
    • Note: System will fallback to local GAN if API is unavailable

Service Levels

  • Primary Service: Uses Stability AI API

    • Requires valid API key and available credits
    • Produces high-quality images at original dimensions
    • Faster processing time

  • Fallback Service: Uses Local GAN

    • Activates automatically if API is unavailable
    • Produces 64x64 pixel images
    • Limited to nature-style variations
    • No API key required

Installation & Setup

Prerequisites

  • NVIDIA AI Workbench installed and configured
  • Docker Desktop running
  • Git installed
  • System Requirements:
    • 8GB RAM minimum
    • 5GB free disk space
  • Software Versions:
    • Docker Desktop 4.x or higher
    • Git 2.x or higher
    • Python 3.8+ (handled by container)

Steps

  1. Open NVIDIA AI Workbench

    • Select "Clone Project"
    • Drop in https://github.com/nhvn/data-aug.git, leave the default path, and click "Clone" Clone Project

  2. Start the Application:

    • In Workbench, click on "Open AugmentAI" (or alternatively, go to Environment > Applications > AugmentAI) Workbench launching AugmentAI
    • The application will open automatically in your default browser

Accessing the Application

The application is available at: http://localhost:10000/projects/data-aug/applications/AugmentAI/

Using the Application

  1. Ensure the app is launched through NVIDIA AI Workbench
  2. Click on 'Augment' in the navigation bar to navigate to the "Upload" page Navigating to Augment Page
  3. Upload images in any of these ways:
  • Click "Load Samples" to use provided test images
  • Drag and drop a folder or individual images
  • Click "Choose Folder" to select files from your computer Upload Methods
  1. Generate augmented images:
  • Click "Augment" to process the images Augment Button
  • Wait for processing to complete
  • Files will automatically download as a ZIP

Note: If the API is unavailable, the system will automatically fall back to a local model (may result in lower quality outputs).

Limitations and Guidelines

  • Supported formats: PNG, JPG, JPEG
  • Maximum 10 files per upload
  • Maximum total upload size: 16MB
  • Maximum image dimensions: 1024x1024 pixels (larger images automatically resized)
  • Folder and file drag-and-drop supported
  • Upload status indication provided

Troubleshooting

If you encounter issues:

  1. Ensure Docker Desktop is running
  2. Verify NVIDIA AI Workbench is properly installed
  3. Check the application logs in Workbench
  4. Ensure your API key is properly configured (if using Stability AI)
  5. Try restarting the application through Workbench

Development Notes

This project is specifically designed for NVIDIA AI Workbench. Key configurations include:

  • NVIDIA AI Workbench integration
  • Stability AI API integration
  • Local GAN model fallback
  • Static file serving optimization

Contact

If you're a judge and need access or have questions, please reach out and I will respond as soon as possible.

Blog Post

Read the full blog post here.