🏋️ Real-Time Squat Analysis Tool (YOLOv8 + PyQt5)

This project provides a real-time squat form analyser that uses YOLOv8 pose estimation inside a PyQt5 GUI.
It measures knee angle and torso angle, gives a traffic-light depth cue, counts reps, and plays an audio beep for every valid squat.

---

📸 Features

• Real-time pose (17 COCO key-points) via YOLOv8.
• Metrics
  • Knee angle (hip – knee – ankle)
  • Torso angle (shoulder – hip vs vertical)
• Traffic light feedback
  • 🟢 ≤ 90 ° (legal depth)
  • 🟠 91 – 110 ° (approaching)
  • 🔴 > 110 ° (not deep enough)
• Rep counter with audio beep.
• Clean GUI (Start / Stop, live feed, overlays, running totals).

---

💡 Best Results

Place the camera side-on to the athlete.
A lateral view captures knee flexion and torso angle most accurately in 2-D.

---

🚀 Getting Started

1 Clone the repository

git clone https://github.com/El3ctricR/squat_tracker_awrc_demo.git
cd squat_tracker_awrc_demo

---

2 Install dependencies

Requires Python 3.8 +. A CUDA-capable GPU is recommended for 30 FPS+. Can be run on CPU be expect reduced performance.

pip install -r requirements.txt

If you prefer manual installation, expand for commands

pip install ultralytics==8.0.20
pip install opencv-python
pip install numpy
pip install PyQt5

Ensure your PyTorch/CUDA versions match your GPU setup — see the Pytorch install guide.

---

🧰 Usage

python squat_gui_app.py

• In the GUI click Start to begin analysis.
• Click Stop to end the session.

During a session you will see:

• Live webcam feed.
• Knee / Torso angles (°).
• Traffic-light depth icon.
• Running rep counter with a beep for each valid squat.

---

📁 File Structure

├── squat_gui_app.py        # Main application (GUI + analysis)
├── yolov8n-pose.pt         # YOLOv8 pose model (download separately if absent)
├── requirements.txt        # Python dependencies
└── README.md               # This documentation

---

📄 License

Released under the MIT License.
See the included LICENSE file for full terms.

---

🙏 Acknowledgements

• 🤖 Ultralytics / YOLOv8 – pose estimation backbone
• 🖼️ OpenCV – real-time computer vision
• 💻 PyQt5 – GUI framework
• 🔉 Windows winsound module – audio feedback for reps

---

📝 Notes & Troubleshooting

Topic	Tip
Audio on macOS/Linux	Replace winsound.Beep() with playsound or pygame.mixer.Sound.
Model download	If yolov8n-pose.pt is missing, download it from the Ultralytics repo or run: from ultralytics import YOLO; YOLO('yolov8n-pose.pt').
Performance	yolov8n-pose.pt offers ~30 FPS on an RTX 3060. Larger models improve accuracy at the cost of speed.
Camera view	Keep the athlete fully in frame and perpendicular to the camera plane for the most reliable depth detection.

---

Happy squatting! Feel free to open issues or PRs if you improve the tool.