VisionTrack is an AI-powered desktop computer vision application that enables real-time object detection, tracking, and counting across images, videos, webcam streams, and batch image collections. Built using YOLOv8, OpenCV, and CustomTkinter, the application combines high-performance computer vision with an intuitive graphical interface, making advanced visual AI accessible to students, researchers, and professionals.
Unlike command-line-based AI tools, VisionTrack provides an interactive desktop experience with multiple detection modes, confidence and class filtering, polygon-based object counting, and on-click object selection. The application is optimized to deliver real-time performance on CPU-only systems while maintaining high detection accuracy.
- 🖼️ Object detection on static images
- 🎥 Real-time object detection and tracking in videos
- 📷 Live webcam object detection
- 🔢 Polygon-based object counting using Regions of Interest (ROI)
- 🖱️ On-click object and region selection
- 📂 Batch processing for 100+ images
- 🎯 Confidence threshold adjustment
- 🏷️ Class-based filtering
- 💾 Export annotated images and videos
- 🌙 Toggleable dark/light theme
- ⚡ Modern desktop interface built with CustomTkinter
- ✅ ~92% image detection accuracy
- 🎥 Real-time webcam detection at approximately 24 FPS
- 💻 Optimized for CPU-only systems (8 GB RAM, no dedicated GPU)
- 📂 Processes 100+ images in approximately 2 minutes
- Language: Python
- AI Model: YOLOv8n (Ultralytics)
- Computer Vision: OpenCV
- GUI: CustomTkinter
- Image Processing: Pillow (PIL)
- Optional AI Module: FER (Emotion Recognition)
- Development Environment: VS Code
Upload an image and automatically detect multiple object classes with confidence filtering and annotated output export.
Track moving objects across video frames, monitor trajectories, and count objects entering predefined polygonal regions.
Perform real-time object detection directly from a live webcam feed for surveillance and monitoring applications.
Select an object or image region with a mouse click to initiate focused detection or object tracking.
Automatically process entire folders containing hundreds of images and save annotated outputs.
- Modern desktop GUI
- Multiple media input sources
- Real-time object tracking
- Polygon ROI-based counting
- Batch image processing
- Interactive object selection
- Confidence and class filtering
- Export annotated outputs
- CPU-optimized inference