Gender Recognition from Images Using Deep Learning

This repository contains a Python-based project for recognizing gender from images using deep learning techniques. The project includes data preprocessing, dataset filtering, and a convolutional neural network model for classification.

Features

• Face alignment using dlib-based landmarks
• Dataset filtering and organization utilities
• Image preprocessing with multiprocessing
• CNN model training with TensorFlow/Keras
• Data augmentation for improved model generalization

Tech Stack

• Python 3
• TensorFlow and Keras for deep learning
• OpenCV for image processing
• dlib for face alignment
• NumPy for numerical operations
• Matplotlib for plotting training history

Getting Started

Prerequisites

• Python 3.6 or higher
• TensorFlow
• dlib
• OpenCV
• NumPy
• Matplotlib

Installation

Clone the repository:

git clone https://github.com/justin-napolitano/Gender-Recognition-from-image.git
cd Gender-Recognition-from-image

Install required packages (preferably in a virtual environment):

pip install tensorflow opencv-python dlib numpy matplotlib

Usage

1. Preprocess images:

python preprocess.py --input_dir path/to/raw_images --output_dir path/to/processed_images --crop_dim 224

2. Filter dataset (adjust parameters as needed):

python FilterDataset.py --input_dir path/to/processed_images --output_dir path/to/filtered_dataset

3. Train the model:

Run the data_generation.py script which contains the model training pipeline. Update directory paths inside the script accordingly.

python data_generation.py

Project Structure

• align_dlib.py: Face alignment module using dlib landmarks (copied from openface project).
• data_generation.py: Defines and trains the CNN model with data augmentation.
• FilterDataset.py: Utilities to filter and organize dataset based on minimum images per class.
• lfw_input.py: TensorFlow queue-based image loader with augmentation (used for batching).
• preprocess.py: Preprocesses images by detecting, aligning, and cropping faces with multiprocessing.
• processed_data/: Directory for processed and filtered dataset images.
• README.md: This file.

Future Work / Roadmap

• Add explicit dataset download and preparation scripts.
• Improve model architecture and hyperparameter tuning.
• Add evaluation scripts and metrics reporting.
• Integrate with a web or mobile interface for real-time gender recognition.
• Expand dataset support and include more diverse data.
• Add unit and integration tests for pipeline components.

Note: Some paths and parameters are hardcoded and should be adapted to your environment.