CoachBuddy

AI-powered basketball coaching system with gimbal tracking, computer vision shot analysis, and real-time feedback.

Built by Rishith Chennupati

---

Overview

CoachBuddy is an open-source basketball training system that uses a phone mounted on an Arduino-controlled gimbal to track a player's movements, analyze shooting form using computer vision, and provide real-time coaching feedback through an iOS app.

How It Works

1. Mount your phone on the gimbal stand at the court
2. Track — The gimbal follows your movement using servo motors and position detection
3. Analyze — Python CV pipeline detects shot attempts, tracks body pose, and analyzes form
4. Feedback — The iOS app shows real-time metrics, shot accuracy, and coaching suggestions

System Architecture

┌─────────────────┐     WiFi/BLE      ┌──────────────────┐
│   iOS App        │◄────────────────►│  Arduino Gimbal   │
│   (SwiftUI)      │                   │  (Servo Control)  │
│                  │                   └──────────────────┘
│  - Live View     │                          │
│  - Shot Metrics  │                     Mounted Phone
│  - Form Score    │                     Camera Feed
│  - History       │                          │
└────────┬─────────┘                          ▼
         │                           ┌──────────────────┐
         │         WebSocket         │  Python CV Server │
         └──────────────────────────►│                  │
                                     │  - MediaPipe Pose│
                                     │  - Shot Detection│
                                     │  - Form Analysis │
                                     │  - Angle Scoring │
                                     └──────────────────┘

Components

1. Gimbal Controller (firmware/)

Arduino-powered 2-axis gimbal that tracks the player and keeps the camera centered. Uses servo motors controlled via PCA9685 with smooth PID-based tracking.

2. Computer Vision (cv/)

Python pipeline using MediaPipe Pose for real-time body tracking. Analyzes shooting form by measuring joint angles at key phases: set point, release, and follow-through. Detects shot attempts and tracks ball trajectory.

3. iOS App (ios-app/)

SwiftUI app that displays the live camera feed, overlays shot metrics, provides coaching feedback, and tracks training history over time.

Hardware Requirements

Component	Purpose	Est. Cost
Arduino Uno R4 WiFi	Gimbal controller	~$30
PCA9685 PWM Driver	Servo control	~$8
2x MG996R Servos	Pan/tilt gimbal	~$20
Phone mount bracket	Camera mounting	~$10
6V 5A Power Supply	Servo power	~$12
3D-printed gimbal frame	Structure	~$8

Total: ~$88

Software Requirements

Python (CV Server)

pip install -r requirements.txt

iOS App

• Xcode 16+
• iOS 18+
• Swift 6

Firmware

• Arduino IDE
• Libraries: WiFiS3, Adafruit_PWMServoDriver, Wire

Quick Start

1. Flash the Gimbal

# Open firmware/gimbal-controller/gimbal-controller.ino in Arduino IDE
# Create credentials.h with your WiFi SSID/password
# Upload to Arduino

2. Start the CV Server

cd cv
python server.py --port 8765

3. Launch the iOS App

open ios-app/CoachBuddy.xcodeproj
# Build and run on your iPhone
# Enter the CV server IP address in Settings

Shot Analysis Metrics

Metric	What It Measures
Elbow Angle	Alignment at set point (ideal: 85-95°)
Release Angle	Wrist angle at ball release (ideal: 45-55°)
Follow-Through	Arm extension after release (ideal: >160°)
Knee Bend	Lower body engagement (ideal: 100-120°)
Balance Score	Shoulder-hip alignment during shot
Arc Consistency	Shot trajectory repeatability

Project Structure

CoachBuddy/
├── firmware/
│   ├── gimbal-controller/     # Arduino gimbal control + PID tracking
│   └── sensor-hub/            # IMU sensor data collection
├── cv/
│   ├── server.py              # WebSocket CV server
│   ├── analysis/
│   │   ├── shot_detector.py   # Detect shot attempts from pose data
│   │   ├── form_analyzer.py   # Score shooting form angles
│   │   └── trajectory.py      # Ball trajectory tracking
│   ├── models/
│   │   └── shot_classifier.py # ML shot type classification
│   └── utils/
│       ├── pose.py            # MediaPipe pose utilities
│       ├── angles.py          # Joint angle calculations
│       └── drawing.py         # Visualization overlays
├── ios-app/
│   └── CoachBuddy/
│       ├── Views/             # SwiftUI views
│       ├── ViewModels/        # MVVM view models
│       ├── Models/            # Data models
│       ├── Services/          # WebSocket, networking
│       └── Components/        # Reusable UI components
├── tests/
│   └── cv/                    # Python CV tests
├── docs/                      # Documentation
└── scripts/                   # Utility scripts

License

MIT License — see LICENSE

Acknowledgments

Built with MediaPipe, OpenCV, Arduino, and SwiftUI. Inspired by wanting to make basketball coaching accessible to everyone.