Age and Gender Prediction

This project utilizes the UTKFace dataset to build a model capable of predicting the age and gender of an individual from their facial image. The dataset comprises over 20,000 images covering a wide range of ages, ethnicities, and genders. The project explores key machine learning concepts including classification tasks, convolutional neural networks (CNN), AI ethics, bias, and model interpretability.

Concepts Explored

• Classification task
• Convolutional Neural Network (CNN)
• AI ethics and bias
• Model interpretability using LIME

Project Goals

• Perform exploratory data analysis (EDA), model building, and evaluation.
• Create a single model that returns both age and gender predictions in a single pass.
• Analyze model performance and identify the best and worst performing samples.
• Utilize LIME for model interpretability with images.
• Evaluate ethical issues and potential biases in the model.

Dataset

The UTKFace dataset consists of facial images labeled with age, gender, and ethnicity. The dataset is highly diverse and includes the following attributes:

• Age: Range from 0 to 116 years.
• Gender: Male (0) and Female (1).
• Ethnicity: White (0), Black (1), Asian (2), Indian (3), Others (4).

Results:

• Age: 4.87 Mean Absolute Error (MAE)

• Gender: 94% F1 Score and confusion matrix

Classification Report:
              precision    recall  f1-score   support

        Male       0.94      0.95      0.94      2214
      Female       0.94      0.93      0.94      2142

    accuracy                           0.94      4356
   macro avg       0.94      0.94      0.94      4356
weighted avg       0.94      0.94      0.94      4356

• Gender: 98% ROC AUC Score

Requirements

The project requires the following libraries and tools:

• Python 3.6+
• PyTorch
• OpenCV
• Matplotlib
• Seaborn
• Scikit-learn
• Pandas
• LIME

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/age-gender-prediction.git

2. Navigate to the project directory:

   cd age-gender-prediction

3. Install the required dependencies:

   pip install -r requirements.txt

Data Preparation

1. Download the UTKFace dataset from the official website.
2. Extract the dataset into the data/utkface_aligned_cropped/UTKFace directory.