Plank.AI - Real-Time Plankton Detection and Tracking
Introduction
Plank.AI is an AI-powered application for real-time plankton detection, tracking, and tallying. It utilizes YOLOv5 for object detection and Norfair for object tracking, integrating seamlessly into a user-friendly PySide6 GUI for visualization and interaction.
Now enhanced with an AI-powered chat assistant, Plank.AI can provide insightful answers and context-aware suggestions based on the plankton species tally metadata.
Features
- Real-time plankton detection using YOLOv5
- Object tracking with Norfair
- AI Chat Assistant integrated into the GUI for data-driven conversations
- Customizable display options (Bounding Boxes, Labels, Object IDs, Plankton Size)
- Camera selection and recording capabilities
- Export detection results to Excel for analysis
- Plankton density calculation based on configurable parameters
- User-friendly GUI built with PySide6
Installation
Prerequisites
Ensure you have Python 3.8+ installed along with the required dependencies.
Install Dependencies
pip install -r requirements.txt
Run the Application
python trial.py
How It Works
- The YOLOv5 model loads pre-trained weights for plankton detection.
- The camera captures live frames, which are processed by the model.
- Norfair tracking maintains object identities across frames.
- The GUI displays real-time bounding boxes, object labels, track IDs, and plankton size.
- The species tally is updated, and density calculations are available.
- The AI Assistant responds to user queries based on the tally's metadata.
- Users can export results to Excel for further analysis.
User Guide
Main Features
- Start, Pause, Stop Detection: Control the object detection process.
- Change Camera Source: Select from available webcam sources.
- Toggle Display Options: Show/hide bounding boxes, object labels, and track IDs.
- Adjust Confidence & Distance Thresholds: Customize detection sensitivity.
- Capture Screenshots & Record Videos: Save detection results.
- Switch Models for different Objectives: supports only 10x and 40x objectives.
- Export Data: Save species tally to an Excel file.
- Plankton Density Calculation: Automatically compute cell density in mL.
- AI Chat Assistant: Ask questions and get real-time answers based on detection data.
Exporting Data
Results can be exported as an Excel file (.xlsx) containing:
- Species Name
- Detection Count
- Cell Density per mL
Technologies Used
- Python 3.8+
- PySide6 (for GUI)
- OpenCV (for image processing)
- YOLOv5 (for object detection)
- Norfair (for tracking)
- Pandas (for data export)
- NumPy (for numerical processing)
- Google Generative AI API (for AI assistant)
Contributors
- Amor Lea T. Palatolon
- Daniel A. Papaya
- Peter Ville C. Carmen
Future Improvements
- Integration with cloud storage for dataset sharing.
- Support for custom YOLO models.
- Performance optimizations for real-time tracking.
- Enhanced post-processing features.
- Web Support Website App.
- Enhanced natural language understanding for the AI Assistant.
- Real-time communication between tally updates and chat responses.
Contact
For inquiries and feedback, contact: peter.carmen0101@gmail.com