VougeVibe: Shop Full Project

VougeVibe is a web-based AI platform that recommends clothing products based on images or selfies uploaded by the user. It leverages YOLO for clothing detection, DeepFace for gender detection, and retrieves similar clothing products from the database for recommendation, and people can use touchless try-on feature (Using VITON-HD).

Video Demnstration Product

Table of Contents

1. Key Features
2. Tech Stack
3. Prerequisites
4. Setup Instructions
5. Future Development

Features

• Image Upload/Selfie: Allows users to upload photos or take selfies for clothing recommendations.
• Clothing Detection: Uses YOLO to detect and categorize clothing items in images.
• Gender Detection: Uses DeepFace to identify the user's gender for more accurate recommendations.

Tech Stack

• Node.js: Backend API and logic.
• MongoDB: Database for storing user and product data.
• YOLO: Deep learning model for clothing detection.
• DeepFace: Gender detection model.
• Firebase: For authentication and real-time data sync.

Prerequisites

Ensure the following are installed on your local machine:

• Node.js (v14 or higher)
• MongoDB (v4 or higher)
• Python (v3.8 or higher)
• Installed YOLO and DeepFace models
• Contact [email protected] for API models.

Setup Instructions

Step 1: Clone the Repository

Clone the repository to your local machine:

git clone https://github.com/HaiDaEmVang/shop_Full.git
cd shop_Full

Step 2: Download and Configure API Models

Run the following command to install dependencies:

# Load Fashion model
path_fashion_model = "../best.pt"  
path_label_model = "../labels.npy"  
path_saved_features = "../saved_features.npy"

Step 3: Add Firebase Configuration

• In the backend folder, paste the fireBaseLog.json file containing your Firebase credentials for user authentication and real-time database operations.
• The best.pt file is the trained YOLO model that detects and categorizes fashion items in images.
• The labels.npy and saved_features.npy file are the featues extraction file of database using DINOv2.
• Put the pre-trained of VITON-HD into VITON-HD/checkpoints/
• You can get the pre-trained file from: [Pre-traind]

Step 4: Start the Application

Create 3 terminals to run the backend, frontend, and admin services: Terminal 1: Start Backend

cd backend
npm install
node index.js

Terminal 2: Start Frontend

cd frontend
npm install
npm run start

Terminal 3: Start Admin Panel

cd admin
npm install
npm run start

Step 5: Start YOLO_API

Step 6: Start Virtual_API

ATTENTION: Run one API at a time

Future Development

VougeVibe will continue to improve with the following enhancements:

• Optimization of Recommendation Speed: Reducing response times to ensure a faster recommendation process.
• Enhanced Fashion Detection Models: Upgrading the YOLO model for more accurate detection and a wider variety of fashion products.
• Increased Product Diversity: Expanding the product database to include more fashion styles, providing more diverse recommendations.
• Improved Image Retrieval: Refining the image retrieval system for more precise recommendations, enhancing feature extraction and cosine similarity for better product matching.
• Enhance segmantation model for VITON-HD