AI-Powered YOLO Annotation Tool

A smart annotation tool that uses YOLOE (YOLO Everything) to accelerate the labeling process. Label one image, and let AI help you label the rest!

Key Features

• Smart Labeling with YOLOE:

  • Draw bounding box on the first image
  • AI automatically detects similar objects in subsequent images
  • Learns from your initial annotation to speed up the labeling process

• Interactive Annotation:

  • Load and browse through images from a folder
  • Draw bounding boxes with click-and-drag
  • Visual prompt-based object detection
  • Review and adjust AI-generated annotations

• Efficient Workflow:

  • Start by annotating just one example
  • AI generates labels for remaining images
  • Review and correct AI predictions if needed
  • Save annotations in YOLO format

Requirements

• Python 3.8+
• PyQt5
• YOLOE dependencies
• OpenCV
• numpy

Installation

1. Clone this repository
2. Install the required packages:

pip install -r requirements.txt

Usage

1. Run the application:

python main.py

2. First Image Annotation:

   • Load your image folder
   • Draw a bounding box around your object of interest
   • Enter the class name for the object
   • This becomes the visual prompt for YOLOE

3. Automatic Labeling:

   • YOLOE uses your first annotation as a reference
   • Automatically detects and labels similar objects
   • Results are shown for review

4. Review and Save:

   • Review AI-generated labels
   • Adjust or correct if needed
   • Save final annotations in YOLO format

YOLO Format

The tool saves annotations in YOLO format:

<object-class> <x-center> <y-center> <width> <height>

• All values are normalized between 0 and 1
• Object class is an integer starting from 0
• (x-center, y-center) represents the center of the bounding box
• (width, height) represents the width and height of the bounding box

Keyboard Shortcuts

• Next Image: Right Arrow
• Previous Image: Left Arrow
• Save: Ctrl+S
• Reset Box: R
• Quit: Q

Roadmap

• Auto-training integration
  • Automatically fine-tune YOLOE on your dataset
  • Improve detection accuracy with more examples
  • Export trained model for inference
• Batch processing capabilities
• Multi-class support in single session
• Export to various formats
• Active learning integration

How It Works

1. Visual Prompting:

   • Your first annotation serves as a visual prompt
   • YOLOE learns what to look for from this example

2. AI Detection:

   • YOLOE processes subsequent images
   • Uses visual similarity to find matching objects
   • Generates bounding box predictions

3. Future: Auto Training:

   • Collect verified annotations
   • Fine-tune YOLOE on your specific dataset
   • Improve detection accuracy over time

Contributing

Contributions are welcome! Please feel free to submit pull requests.