Project Deep Dive

Golf Swing AI

A computer vision system that brings professional-grade biomechanical analysis to every golfer's pocket.

Overview

Golf Swing AI is a full-stack machine learning application that captures video of a golfer's swing and delivers real-time pose analysis and correction feedback. It bridges the gap between expensive professional coaching and the self-taught amateur golfer.

The system processes video frames through a MediaPipe pipeline to extract 3D skeleton coordinates, computes joint angles, classifies swing phase, and compares biomechanical signatures against a dataset of professional swings to generate prioritized correction recommendations.

StatusLive & deployed
TechPython · MediaPipe · TF
InterfaceStreamlit Web App
TypeComputer Vision / ML

The Problem

  • Professional swing coaching costs $100–$300 per session
  • Amateur golfers lack access to biomechanical feedback
  • Video review requires trained professionals to interpret
  • Generic tips don't address individual mechanics

The Solution

  • Free, instant AI analysis accessible from any device
  • Pose estimation extracts precise joint angles automatically
  • ML models compare against professional swing patterns
  • Personalized, prioritized corrections based on skill level
Features

What it does

Real-time Pose Detection

MediaPipe detects 33 body landmarks at 30+ FPS for live swing analysis.

Instant Feedback

Biomechanical scores and correction tips appear within milliseconds of each swing.

Custom ML Models

Fine-tuned classifiers identify swing phases: address, backswing, impact, follow-through.

Skill-Level Aware

Adaptive feedback calibrated to beginner, intermediate, and advanced golfers.

Tech Stack

Built with

Vision & AI

  • MediaPipe
  • TensorFlow / Keras
  • OpenCV
  • NumPy
  • scikit-learn

App Layer

  • Python 3.11
  • Streamlit
  • FastAPI (optional backend)

Data & Analytics

  • Pandas
  • Matplotlib
  • Seaborn

Deployment

  • Streamlit Cloud
  • Docker
  • GitHub Actions
My Contribution

Built end-to-end by Aneesh

Designed the full ML pipeline from data collection to deployment

Built and tuned the pose estimation + swing classification models

Developed the Streamlit web interface and deployed to cloud

Try the App