Emotion, Age & Gender Detection using Deep Learning

This project performs real-time facial emotion, age, and gender detection using Convolutional Neural Networks (CNNs) with a webcam feed. It leverages the CK+ dataset for emotion recognition and custom CNNs trained for predicting age and gender.

---

Project Structure

emotion-age-gender-detection/

│

├── emotion_finder.py # Main python file for real-time detection script

├── emotion_model.keras # Trained emotion detection model

├── gender_model.keras # Trained gender classification model

├── age_model.keras # Trained age regression model

├── haarcascade_frontalface_default.xml # Haar Cascade for face detection

├── requirements.txt # Python dependencies

├── emotion_finder.ipynb # Main jupyter notebook for real-time detection script

└── README.md # Project documentation

---

Model Details

Emotion Detection

• Trained on: CK+48 (resized grayscale 48x48)
• Output: 7 emotion classes (angry, happy, sad, fear, disgust, surprised, contempt)
• Activation: Softmax

Gender Detection

• Trained on: UTKFace subset
• Output: Binary classification (Male, Female)
• Activation: Sigmoid

Age Prediction

• Trained on: UTKFace
• Output: Single regression output (0–100)
• Activation: Linear

---

Running the Real-Time Detector

Install Requirements

python version 3.10.9 (Recommended) or lower required

pip install -r requirements.txt

Run the script

python emotion_detector.py

To quit the webcam

Press q to quit the webcam window.

Output Sample

When you run the webcam script, each detected face is annotated with:

Emotion label (top of the face)

Gender (below the face on the left)

Predicted age (below the face on the right)

Datasets used

1)https://www.kaggle.com/datasets/shawon10/ckplus [for expression recognition]

2)https://www.kaggle.com/datasets/moritzm00/utkface-cropped [for age and gender recognition]