DLBAIPEAI – Age, Gender, and Expression Recognition on Edge Devices

An edge-deployable deep learning application that predicts Age Group, Gender, and Facial Expression in real time using a mobile device. Built using PyTorch, EfficientNetV2, and deployed via Android Studio for fast and private offline inference.

---

📱 Features

• 📸 Real-time face detection and recognition on mobile devices
• 🧠 Multi-head deep learning model with shared EfficientNetV2 backbone
• 🎭 Recognizes:
  • Age Group: Baby, Child, Teen, Adult, Elderly
  • Gender: Male, Female
  • Expression: Happy, Sad, Neutral, Angry, Surprised,
• ⚡ Optimized for edge devices (runs offline, no cloud dependency)

---

🧠 Model Architecture

A multi-task classification model with shared representation:

• Backbone: EfficientNetV2 for feature extraction
• Classification Heads:
  • Age Group (5 classes)
  • Gender (2 classes)
  • Emotion (5 classes)

This approach allows efficient learning from shared facial features while outputting multiple predictions.

---

🔧 Requirements

🔹 Python Side

• Python 3.8+
• PyTorch
• TorchVision
• NumPy
• OpenCV

🔹 Android Side

• Android Studio (Arctic Fox or newer)
• Java 8+
• Gradle
• Android device or emulator

---

Deploy on Android

• Open DLBAIPEAI in Android Studio.
• Place the .ptl model inside the app/src/main/assets/ folder.
• Connect your Android phone or emulator and run the app.

📊 Datasets Used

• Age Classification: Face age
• Gender: Biggest Gender Face Recognition Dataset
• Emotion: Facial Emotion Dataset

---

👩‍💻 Author

Fotimakhon Gulomova

• 🎓 AI Student at IU International University
• 🌐 GitHub
• 🔗 LinkedIn
• 🔗 Kaggle

📜 License

This project is licensed under the MIT License. Feel free to use, modify, and contribute.