SpinMaster is your ultimate virtual table tennis analyst — a full-stack AI application that breaks down match footage, identifies gameplay weaknesses, and recommends personalized training modules to help players level up their performance like the pros.
- 🎥 Video Upload Interface — Upload match videos directly through a sleek web interface.
- 🤖 Automated Pose Detection — Detects elbow angles and arm orientation to classify forehand and backhand shots.
- 📊 Match Analysis Engine — Calculates shot success rates and determines weak areas.
- 🧠 TTNet Integration (Modular) — Built to support advanced table tennis models like TTNet for precise shot classification.
- 📈 Dynamic Leaderboards — Rank top players using comic-inspired UI.
- 🎓 Smart Training Suggestions — Recommends YouTube tutorials based on your weaknesses.
- 🔐 User Authentication — Register and log in to track your progress (with session support).
- 🌐 Flask-based Local Server — Lightweight, local-first deployment with modern UI/UX.
Utilizes MediaPipe pose landmarks to:
- Extract joint positions (specifically elbow angles)
- Identify the shot type (forehand/backhand) using angle deltas
- Timestamp each frame
- Track shot success based on rally outcome
- Export structured
.csvdata for analysis
Reads the generated CSV and:
- Computes per-player statistics
- Calculates shot-wise success ratios
- Determines point and match winners
- Feeds performance summary to front-end (forehand/backhand strengths)
A simple entry point that merges the above scripts for batch processing or CLI testing.
Your all-in-one Flask server:
- Manages video uploads
- Runs the full analysis pipeline
- Displays real-time feedback
- Handles user login/registration
- Renders leaderboard and training module pages
SpinMaster is built with modularity in mind. While it runs a lightweight elbow-angle classification model by default, it also supports powerful integrations like:
A dedicated table tennis recognition model trained on extensive datasets to detect:
- Forehand / Backhand
- Serve / Smash / Topspin
- Ball location & player segmentation
🧪 We're currently integrating TTNet as an alternate backend for precision-grade analysis using deep learning.
- Pastel & Neon Dark Mode 🎨
- Glassmorphism cards and overlays
- Background video or image with motivational quotes
- Embedded YouTube training cards
- Responsive layout (desktop-first)
- Clone the repository
git clone https://github.com/yourusername/spinmaster.git
cd spinmaster-
Install dependencies
pip install -r requirements.txt
-
Launch the Flask app
python app.py
-
Open in browser:
http://127.0.0.1:5000/
SpinMaster/
├── app.py
├── pose_to_csv.py
├── csv_analyzer.py
├── analyze_match.py
├── users.json
├── uploads/
├── static/
│ ├── styles.css
│ ├── login-icon.png
│ └── tt.jpg / tt.mp4
└── templates/
├── index.html
├── login.html
├── leaderboard.html
└── training.html
Abraash Syed Futurist | AI Developer | Table Tennis Enthusiast 🏓 GitHub • LinkedIn
Special thanks to the incredible individuals and organizations whose work and contributions made SpinMaster possible:
Lab OSAI — For providing the table tennis datasets used to train and test TTNet Nguyen Mau Dzung — For detailed guidance and model structure of the TTNet architecture MediaPipe — For enabling seamless pose tracking and landmark detection Nicholas Renotte — For insightful tutorials and guidance on object recognition using the Roboflow platform WTT (World Table Tennis) — For global player rankings and match insights And all other open-source contributors, documentation authors, and creators whose tools and knowledge supported the development of this project.
Without this collective foundation, SpinMaster would not exist.
MIT License © 2025 Abraash Syed Feel free to use, contribute, and help SpinMaster grow!
“Master the spin, master the game.”