Facial Gender Classification

Project Overview

This project aims to develop a machine learning model for classifying gender based on facial images. The dataset used for this project is the UTKFace dataset, which contains a large collection of images labeled with gender, age, and ethnicity. The primary goal of this project is to build a model that can accurately distinguish between male and female faces.

Dataset

The dataset is sourced from Kaggle and includes over 20,000 images of faces with the following attributes:

• Gender: Male or Female
• Age: From 0 to 116 years
• Ethnicity: White, Black, Asian, Indian, and Others

The dataset can be found here.

Model Architecture

The model is built using convolutional neural networks (CNNs) to capture the spatial hierarchies in the facial images. The architecture includes:

• Input Layer: Accepts 48x48 pixel grayscale images.
• Convolutional Layers: Extracts features through a series of convolutional operations.
• Pooling Layers: Reduces the spatial dimensions of the feature maps.
• Fully Connected Layers: Combines the extracted features to predict the gender.

Implementation

The implementation is done in Python using the following libraries:

• TensorFlow/Keras for building and training the neural network.
• OpenCV for image preprocessing.
• NumPy and Pandas for data manipulation.

Training and Evaluation

The model is trained on 80% of the dataset, with the remaining 20% used for validation and testing. Various data augmentation techniques, such as rotation, flipping, and scaling, are applied to increase the robustness of the model.

Key Metrics:

• Accuracy: The primary metric for evaluating the model's performance.
• Precision, Recall, and F1-Score: Additional metrics for assessing the model's classification abilities.

Results

The model achieves an accuracy of over 78.73% on the validation set, demonstrating its effectiveness in gender classification. Further improvements could involve fine-tuning the model architecture or experimenting with more complex neural networks.

Usage

To use the model, follow these steps:

1. Clone the repository.
2. Install the necessary dependencies, like tensorflow, opencv-python
3. load the models using keras.model.load_model

Contributing

Contributions are welcome! Please fork the repository and submit a pull request.

License

This project is licensed under the MIT License - see the LICENSE.md file for details.