🍍 PineCount AI

An advanced computer vision system for detecting and counting pineapple blooms in drone-captured imagery. This application combines YOLOv8 object detection with explainable AI techniques to provide insights into the model's decision-making process through an intuitive Streamlit interface.

✨ Features

• 🖼️ Image Processing: Upload and preprocess drone images for optimal detection
• 🔍 Bloom Detection: Utilize YOLOv8 to accurately identify and count pineapple blooms
• 🧠 Explainable AI (XAI): Visualize and understand the model's decisions with heatmaps and feature importance
• 📊 Performance Monitoring: Track and visualize model performance metrics over time
• 🌐 Interactive Interface: User-friendly Streamlit interface for all functionalities

🏗️ Project Structure

PineCount-AI/
├── src/
│   ├── main.py          # Main Streamlit application
│   ├── model.py         # Functions to load, preprocess, and run the model
│   ├── xai.py           # Functions to generate model explanations (XAI)
│   └── monitoring.py    # Functions to log and visualize performance metrics
│
├── weights/
│   └── weights.pt       # Pre-trained YOLOv8 model weights
│
├── logs/                # Directory for performance logs (created at runtime)
│
├── assets/              # Images and other static assets
│   └── pinecount-logo.svg  # Project logo
│
├── requirements.txt     # List of necessary libraries
└── README.md            # This file

🚀 Installation

1. Clone this repository:

   git clone https://github.com/AndresACV/PineCount-AI.git
   cd PineCount-AI

2. Create and activate a virtual environment:

   # On Windows
   python -m venv venv
   venv\Scripts\activate
   
   # On macOS/Linux
   python -m venv venv
   source venv/bin/activate

3. Install the required dependencies:

   pip install -r requirements.txt

🎮 Usage

1. Start the Streamlit application:

   streamlit run src/main.py

2. Use the web interface to:

   • Upload drone images of pineapple fields
   • Run the detection model to count pineapple blooms
   • View visualizations of detected blooms
   • Generate and explore model explanations
   • Monitor model performance metrics

🧪 Model Information

The pineapple bloom detection model is implemented using YOLOv8, a state-of-the-art object detection framework. The pre-trained weights are stored in the weights/weights.pt file.

🔍 Explainability

The XAI features include:

• Confidence score visualization for each detected bloom
• Attention heatmaps showing which parts of the image the model focuses on
• Detailed statistics about detections including confidence levels and size distributions

📊 Monitoring

Performance metrics tracked include:

• Inference time
• Bloom count distribution
• Confidence levels
• Summary statistics

📦 Requirements

• Python 3.9+
• PyTorch 2.0+
• Ultralytics (YOLOv8)
• Streamlit
• OpenCV
• SHAP
• Other dependencies listed in requirements.txt

📄 License

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

👥 Contributors

Andrés Calvo - GitHub Profile

🙏 Acknowledgements

• Ultralytics for the YOLOv8 implementation
• Streamlit for the web application framework
• SHAP for model explainability tools