👣 AI-Gait-Visualizer

---

📍 Overview

AI-Gait-Visualizer is an interactive application that extracts human pose keypoints and gait metrics from short walking videos.
It uses YOLOv8-Pose for keypoint detection and a lightweight gait-analysis pipeline to estimate stride length proxies and cadence. This application is built with Ultralytics YOLOv8, OpenCV, and Gradio. It allows you to upload a walking video and instantly visualize gait characteristics. You do not need to install anything or write a single line of code. Just open it, upload your video, and explore how AI interprets human motion.

---

🚀 Live Demo

🟢 Try it directly here:
👉 https://huggingface.co/spaces/rmehmood786/AI-Gait-Visualizer

Upload a walking video → click Run Analysis → view the annotated video and download CSV metrics.

---

⚙️ Features

• 🎯 Pose Estimation: Real-time body keypoints using YOLOv8-Pose.
• 🦶 Gait Metrics: Calculates stride proxy, step duration, and approximate cadence.
• 📊 Downloadable Results: Exports a CSV file with per-frame gait data.
• 💡 Interactive Interface: Built with Gradio for smooth use.
• ☁️ Cloud-Ready: Runs directly on Hugging Face Spaces.

---

🧩 Tech Stack

Component	Description
Model	YOLOv8-Pose (Ultralytics)
Frameworks	Gradio, OpenCV, NumPy, SciPy, Pandas
Deployment	Hugging Face Spaces
Language	Python 3.10+

---

🗂️ Project Structure

---

🚀 Live Demo

🟢 Try it directly here:
👉 https://huggingface.co/spaces/rmehmood786/AI-Gait-Visualizer

Upload a walking video → click Run Analysis → view the annotated video and download CSV metrics.

---

⚙️ Features

• 🎯 Pose Estimation: Real-time body keypoints using YOLOv8-Pose.
• 🦶 Gait Metrics: Calculates stride proxy, step duration, and approximate cadence.
• 📊 Downloadable Results: Exports a CSV file with per-frame gait data.
• 💡 Interactive Interface: Built with Gradio for smooth use.
• ☁️ Cloud-Ready: Runs directly on Hugging Face Spaces.

---

🧩 Tech Stack

Component	Description
Model	YOLOv8-Pose (Ultralytics)
Frameworks	Gradio, OpenCV, NumPy, SciPy, Pandas
Deployment	Hugging Face Spaces
Language	Python 3.10+

---

🗂️ Project Structure

AI-Gait-Visualizer/
│
├── app/
│   └── app.py              # Gradio app entry point
│
├── src/
│   └── gait_viz.py         # Gait extraction + pose processing
│
├── requirements.txt         # Dependencies
├── app_file                 # tells Hugging Face to run app/app.py
└── README.md

---

🧠 How It Works

1. The uploaded video is read frame-by-frame using OpenCV.
2. YOLOv8-Pose detects body keypoints per frame.
3. Gait parameters (stride proxy, step duration, cadence) are computed from keypoint motion.
4. Processed frames are written back into an annotated video.
5. The app displays results and offers a downloadable CSV.

---

🖥️ Run Locally

git clone https://github.com/rmehmood786/AI-Gait-Visualizer.git
cd AI-Gait-Visualizer
pip install -r requirements.txt
python app/app.py

Then open the local Gradio link in your browser.

---

📦 Requirements

gradio>=4.0.0
ultralytics>=8.2.0
opencv-python>=4.10.0
numpy>=1.24
pandas>=2.1
scipy>=1.11
matplotlib>=3.8
tqdm
rich

---

🎬 Example Output

Input	Output

---

💡 Ideas to Extend or Experiment With

• Integrate pose estimation with gait recognition for identity prediction.
• Visualize motion trajectories in 3D using Open3D or Matplotlib.
• Add real-time gait visualization using webcam input.
• Try comparing different YOLO variants (v8, v10) for accuracy vs. speed.
• Use temporal smoothing or Kalman filtering for stable keypoints.
• Build a web-based dashboard for visualizing stride metrics interactively.

---

👤 Author

Rashid Mehmood
AI Researcher — Computer Vision & Gait Analysis

• 🌐 GitHub
• 🤗 Hugging Face Space

---

🪪 License

Released under the MIT License.
You are free to use, modify, and share with proper attribution.

---

⭐ Acknowledgements

Built using:

• Ultralytics YOLOv8
• Gradio