ShapeStyle

Overview

An AI-powered web application that analyzes user photos to determine face shapes and provides personalized hairstyle recommendations. Using deep learning and transfer learning techniques, the system classifies faces into five categories (Heart, Oblong, Oval, Round, Square) and suggests optimal hairstyles for each shape.

Purpose

To help users find the most suitable hairstyle for their face shape through automated facial analysis. The application aims to accelerate decision-making in hairstyle selection and ensure users make the right choice by matching their facial features with predefined hairstyle recommendations.

Scope

Technology Stack:

• Deep Learning: VGGFace, Keras 2.2.4, TensorFlow 1.14.0
• Face Detection: MTCNN
• Web Development: Python Flask, HTML/CSS, JavaScript
• Data Processing: NumPy, OpenCV, joblib, pickle
• Transfer Learning: Pre-trained VGG16 architecture

Implementation

Project Structure:

ShapeStyle/
├── model/
│   ├── face_shape_model_vgg16_rgb.h5    # Trained model
│   ├── evaluation_results/               # Performance metrics and graphs
│   ├── input_imgs/                       # Test images
│   ├── output_results/                   # Prediction results
│   ├── face_shape_trainer.py                    # Model training script
│   ├── face_shape_predictor.py                     # Model testing script
│   └── face_shape_evaluator.py           # Model evaluation script
├── preprocessing/
│   ├── preprocessing.py                  # Data preprocessing script
│   └── preprocessing_control.py          # Data validation script
├── hair_style_recommender/
│   ├── backend/
│   │   └── api.py                       # Flask API server
│   └── frontend/
│       ├── face_shape_classifier.html   # Main interface
│       ├── recommended_styles.html      # Hairstyle recommendations
│       └── [face_shape_folders]/        # Hairstyle images by face shape
└── requirements.txt

Model Performance:

• Accuracy: 89.60%
• Precision (Weighted): 89.62%
• Recall (Weighted): 89.60%
• F1-Score (Weighted): 89.59%
• AUC Score (Weighted): 98.84%

Classification Report:

Face Shape	Precision	Recall	F1-Score	Support
Heart	0.92	0.93	0.92	200
Oblong	0.93	0.93	0.93	200
Oval	0.89	0.84	0.87	200
Round	0.86	0.89	0.87	200
Square	0.89	0.90	0.89	200

Screenshots

System Workflow

Model Performance Visualizations

Dataset:

• 5,000 celebrity images categorized by face shapes
• 1,000 images per category (Heart, Oblong, Oval, Round, Square)
• Training set: 800 images per category
• Test set: 200 images per category