This project is a fully-featured GUI application for real-time object detection and tracking using YOLOv8. It lets users load videos, preview detections live, customize bounding box colors, and export annotated videos—complete with original audio.
- ✅ YOLOv8 Object Detection and Tracking
- 🖼️ Tkinter GUI with Live Video Preview
- 📂 Drag-and-Drop or File Picker for Input Videos
- 🎨 Customizable Bounding Box Colors (RGB)
- 🔊 Audio Preserved Using FFmpeg (if installed)
- 📊 Progress Bar and Console Logging
- ⚙️ Threaded Video Processing
- 🧠 Automatic CUDA Detection and Usage (if available)
- 📁 Auto-Saved Output in Timestamped Folders
- 👀 Updated UI: Totally inspired by our website aesthetics.
- 🔲 Interactive Region of Interest (ROI): Draw a custom bounding box on the first frame of your video to restrict YOLOv8 tracking to a specific area. This dramatically reduces processing overhead and focuses your results.
- 📱 Portrait Video Correction: Automatically detects video rotation metadata and rotates mobile portrait videos to their correct orientation before processing.
git clone https://github.com/ProjectGlyphMotion/GUI && cd GUIpip install -r requirements.txtOn Linux, you may also need:
sudo apt install python3-pil.imagetkEnsure FFmpeg is installed and accessible from the command line.
By default, pip install torch installs the CPU-only version. The app will still work on CPU, but tracking will be significantly slower.
To enable GPU acceleration, install PyTorch with CUDA from the official PyTorch site based on your setup:
| CUDA Version | Install Command |
|---|---|
| CUDA 11.8 | pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118 |
| CUDA 12.1 | pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121 |
| CUDA 12.4 | pip install torch torchvision --index-url https://download.pytorch.org/whl/cu124 |
| CPU Only | pip install torch torchvision (default) |
Not sure which CUDA version you have? Run
nvcc --versionornvidia-smiin your terminal.
⚠️ CUDA can be a pain in the 🍑 if you have a 30 or 40 series card — here is a FIX
Verify GPU is detected:
python -c "import torch; print(torch.cuda.is_available())"python3 GUI.py- Browse or drag a
.mp4,.avi, or.movfile. - Select the desired bounding box color (RGB or via color picker).
- Optionally you can also use the Region of Interest (ROI)
- Click ▶ Run Tracker.
- Note: This looks kinda distorted because of the conversion.
- Format:
<original_filename>_tracked.mp4 - Saved under:
output/YYYYMMDD-HHMMSS/ - If FFmpeg is available, original audio is preserved.
✅ Example Result
MIT License © 2025 Sayan Sarkar & Shitij Halder
Made with love by Sayan and Shitij
This project is based on the Ultralytics YOLOv8, an acclaimed real-time object detection and image segmentation model.