Vision Studio Domain

A.R.I.E.S — Universal Virtual Try-On & Fit Engine

Architecting an end-to-end computer vision ecosystem: Manifest V3 client capture, asynchronous FastAPI data pipelines, and MediaPipe human body frame measurement.

From Product Page to Fit Result

The path a garment takes through the platform, end to end.

1. Capture

The Manifest V3 extension injects action hooks into the product page.

2. Store

Metadata and images land in the PostgreSQL wardrobe store.

3. Serve

Async FastAPI endpoints, secured with JWT, sync carts and libraries.

4. Measure

MediaPipe Pose maps body keypoints from a webcam frame.

5. Fit

Shoulder width and torso height drive the fit result, locally.

Cross-Platform E-Commerce Ingestion

The platform bridges external e-commerce web storefronts with a centralized virtual dressing room using a Manifest V3 browser extension. It dynamically injects action hooks into any product page to extract product metadata and image assets directly into a PostgreSQL wardrobe store.

The backend services are engineered around an asynchronous FastAPI framework, exposing scalable REST API endpoints secured via JWT authentication for session management, cart syncing, and user garment libraries.

Fig 1: Manifest V3 injection pipeline and FastAPI asynchronous architecture.

Fig 2: Standalone OpenCV & MediaPipe keypoint detection for body measurement.

Real-Time Anthropometric Fit Engine

At the core of the fit analysis system is a Python inference engine leveraging MediaPipe Pose tracking and NumPy to compute real-time anthropometric keypoints directly from standard webcam frames.

Keypoint Estimation: Maps body landmarks to calculate shoulder width, torso height, and frame scaling.
Snapshot Capture: Custom OpenCV triggers allow instant visual measurement logging.
Local Execution: Operates locally on edge devices without requiring server round-trips.

Dashboard & Generative Roadmap

Next.js 14 Dashboard

Interactive App Router dashboard with Tailwind styling and live webcam canvas controls.

PostgreSQL & SQLAlchemy

Async database schema managing user profiles, custom wardrobes, and fit history.

IDM-VTON Diffusion

Planned GPU inference server integration for neural garment warping and synthesis.

Semantic Parsing

SCHP / DeepLabV3+ segmentation pipeline for precise clothing mask isolation.

Fig 3: Next.js frontend interface coupled with real-time pose analytics.

Quickstart & Local Setup Guide

1. Repository & FastAPI Backend

# Clone repository
git clone https://github.com/siddarthgulimi/aries_tryon-.git
cd aries_tryon-/universal-tryon/backend

# Create virtual environment & dependencies
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# Database setup & server execution
createdb tryondb
cp .env.example .env
uvicorn main:app --reload --host 0.0.0.0 --port 8000

2. Next.js 14 Dashboard

# Navigate to frontend folder
cd ../frontend

# Install node dependencies
npm install
cp .env.local.example .env.local

# Run development server
npm run dev

Access dashboard at http://localhost:3000.

3. Manifest V3 Extension

Open Chrome and navigate to chrome://extensions.
Toggle Developer Mode on in the top-right corner.
Click Load unpacked and select the universal-tryon/extension/ folder.
Update apiBase in `content.js`, `background.js`, and `popup.js` to target your local API endpoint.

4. Standalone Fit Engine CLI

# Navigate to fit-engine
cd ../fit-engine
pip install -r requirements.txt

# Run pose measurement CLI
python fit_engine.py --height-cm 175 --garment garment.example.json

Press s in the OpenCV window to capture a measurement snapshot or q to quit.

Try It Yourself

Run the project live in your browser, or install the open-source build on your own machine.

Live demo

Run it in your browser

The A.R.I.E.S dashboard runs here with an in-browser port of its API: the imported cart, the Dressing Room with live size switching and fit status, and the Digital Wardrobe. Sign in with demo@aries.app / demo1234.

RUN DEMO

Product capture from real stores needs the browser extension; the demo cart is pre-filled with sample garments.

Open source

Install it on your machine

Run the FastAPI backend with PostgreSQL, the Next.js dashboard, the Manifest V3 extension that captures products from any store, and the MediaPipe fit engine.

INSTALL THE PROJECT (OPEN SOURCE) Source code on GitHub

Explore the Source Code

Access the complete codebase containing the Manifest V3 browser extension, FastAPI server, MediaPipe fit engine CLI, and Next.js frontend application.

View Repository on GitHub