Accessible AI Boxing Coach: Style-Adaptive Form & Combo Trainer

Tagline

An AI-powered, style-adaptive boxing coach that brings elite personal training and form analysis to everyone.


1. The Vision & Mission

Professional boxing training and personal coaching are expensive and inaccessible to most. Our mission is to democratize elite sports training by providing an intelligent, personalized, and accessible coaching tool using affordable hardware ($25 Arduino kit + laptop webcam).

Unlike generic fitness apps that bark one-size-fits-all commands, our system adapts completely to the individual user's body dimensions, stance, and natural fighting style.


2. System Architecture & How It Works

                        +---------------------------------------+
 Webcam --------------> | Laptop (Python Backend, FastAPI, UI)  |<----- Web UI
 (Pose X/Y, Onset)      | - MediaPipe Body & Reach Calibration  |       - Signature Style
                        | - Combo & Flaw Detection Engine       |       - Session Overview
                        +-------------------+-------------------+       - Goal System
                                            | USB Serial, 115200 baud
                                     +------v------+
                                     | Arduino Uno |
                                     +--+--+--+--+-+
              Ultrasonic (Z Depth) ----+  |  |  +---- OLED Display
                                          |  +------ LEDs (Green/Yellow/Red)

Key Components:

  1. Laptop Vision (MediaPipe): Tracks upper-body joints, stance (orthodox/southpaw), and movement timing.
  2. Arduino Hardware Unit: Features an HC-SR04 ultrasonic sensor for precise depth/extension tracking, 3 feedback LEDs (Green/Yellow/Red), and an I2C OLED display for quick stats.
  3. Smart Integration: The Arduino handles physical I/O and timing, while the laptop runs the heavy AI, computer vision, and workout logic.

3. Core Features (Step-by-Step User Journey)

  1. Accessible Onboarding & Clothing Tip: Welcomes the user and gently advises avoiding overly baggy winter clothing for optimal vision tracking.
  2. Non-Judgmental Mechanical Calibration: Instead of judging body types, the system quickly scans the user to map their exact reach, height, and stance to dynamically calibrate the ultrasonic sensor's distance gate.
  3. Style-Aware Combo & Flaw Detection: Guides users through targeted combos. Uses computer vision and ultrasonic depth tracking to catch mechanical flaws (like dropped hands or short-armed punches) in real-time, flashing physical LEDs for instant feedback.
  4. Your "Signature Style": Tracks the user's natural rhythm and preferences to discover and display their unique fighting style and most-frequently used combo.
  5. Session Overview & Goal Tracker: At the end of every workout, the app generates a comprehensive breakdown showing achievements, next steps, a review of session goals set vs. achieved, and an option to set new goals for the next workout.
Share this project:

Updates

Submission history