AI biomechanics for accessible running analysis

Runalyst

AI-powered running form analysis from a single smartphone video.

Overview

Running form analysis without lab equipment.

Runalyst makes running analysis accessible by reconstructing a 3D body model from short running videos, calculating biomechanical metrics, and turning those results into clear performance and injury-prevention guidance.

RunnersCoachesFitness enthusiastsSports science teams
Features

From capture to coaching, each step is designed for everyday runners.

Runalyst combines mobile-first capture, cloud inference, biomechanical analysis, and personalized feedback in a single product flow.

01

Video Capture & Quality Guidance

Users record short running videos in the mobile app. Preflight checks evaluate duration and framing before upload.

02

3D Reconstruction Pipeline

Uploaded videos are processed through an SMPL inference pipeline using NLF baseline powered by PyTorch.

03

Biomechanical Metrics

The system analyzes stride length, cadence, ground contact time, swing phase duration, knee motion, posture, and asymmetry.

04

Actionable Coaching

Runalyst prioritizes detected form issues and suggests drills, exercises, and technique modifications tailored to each user.

05

Progress Tracking

A dashboard presents current analysis results and historical trends across multiple video analyses.

06

Report Q&A

A simple chatbot-like interface answers common questions about the user's analysis report with templated responses.

Technical

A secure cloud pipeline for video, 3D pose, and movement analytics.

The architecture separates the mobile client, API, inference workers, storage, queues, and biomechanical analysis services so each stage can scale independently.

React NativeFastAPIPyTorchSMPLNLFSupabase StorageAWS sqs
1

The mobile client uploads video through the API.

2

The API creates a job, stores metadata, and enqueues SMPL inference.

3

The inference server retrieves the video, estimates pose, and saves canonical outputs.

4

A biomechanical analysis server computes gait, knee, posture and asymmetry metrics.

5

The app fetches results and visualizes them in a dashboard.

Team

Built by a METU CENG senior project team Runalyst.

The team works across mobile development, backend systems, ML inference, analysis pipelines, and product experience.

Academic advisor: Emre AkbasExternal advisor: Muhammed Kocabas