Fromsa Teshome Negasa

Fromsa Teshome Negasa

Masters Student at EMJMD PSRS

Face Analysis System: Emotion, Age, and Gender Estimation

Overview

This project is a multi-task face analysis system that combines three biometric capabilities into a single PyQt5-based GUI application:

  1. Emotion Estimation
  2. Age Estimation
  3. Gender Estimation

Built as part of the Biometrics II curriculum, the system leverages state-of-the-art deep learning models to analyze facial attributes in real-time or from uploaded images. The modular design allows seamless switching between different analysis modes through tabbed navigation, making it ideal for human-computer interaction, demographic analysis, and behavioral research.

Emotion Estimation Tab
Emotion Estimation Tab
Age Estimation Tab
Age Estimation Tab

Key Features

Unified Application Architecture

  • Single Interface: Three specialized tabs for emotion, age, and gender estimation
  • Real-Time Processing: Webcam integration for live analysis (10-15 FPS)
  • Batch Image Analysis: Process multiple faces in uploaded images
  • Visual Annotation: Color-coded bounding boxes with attribute labels
  • Model Switching: Select between lightweight and high-accuracy models per task

Emotion Estimation Tab

  • 🎭 7 Emotion Classes: Angry, Disgust, Fear, Happy, Sad, Surprise, Neutral
  • 🤖 Models: MobileNetV2 (85% accuracy) + Custom Sequential CNN (78% accuracy)
  • 📈 Confidence Display: Real-time emotion probability distribution

Age Estimation Tab

  • 👶👴 Age Range: 0-116 years (UTKFace dataset)
  • 🤖 Models: ResNet50 (MAE=6.2 years) + Custom CNN (MAE=8.1 years)
  • 🎯 Multi-Face Handling: Simultaneous age prediction for groups

Gender Estimation Tab

  • 👨👩 Binary Classification: Male/Female with probability scores
  • 🤖 Models: ResNet50 (94% accuracy) + Custom CNN (89% accuracy)
  • ⚡ Optimized Inference: <150ms per face on CPU

Technical Details

Shared Components

  • Face Detection: OpenCV Haar Cascades

Common Infrastructure

  • GUI Framework: PyQt5
  • Backend: TensorFlow 2.x
  • Webcam Processing: Multi-threaded
  • Model Management System

Demo

Live Demo of Face Emotion Estimation
Live Demo of Face Age Estimation
Live Demo of Face Gender Estimation

Impact & Applications

This integrated system demonstrates how multiple biometric tasks can be efficiently combined into a single application:

  • Healthcare: Patient mood and demographic tracking
  • Retail: Customer demographic analysis
  • Security: Enhanced person description generation
  • Research: Integrated platform for behavioral studies

Future Improvements

  • Multi-Task Learning: Single model for all attributes
  • Edge Deployment: TensorFlow Lite for mobile devices
  • 3D Face Modeling: Improved angle robustness
  • Privacy Features: On-device processing only