Face Detection and Classification

This repository contains a reimplementation of the FaceNet model for face detection and classification. The model is trained to predict the name of a person appearing in an image.

Table of Contents

• Face Detection and Classification
  • Table of Contents
  • Introduction
  • Installation
  • Dataset
  • Usage
  • Training
  • Results
  • Contributing
  • License

Introduction

The Face Detection and Classification project aims to detect and classify faces in images to predict the name of the person in each detected face. This repository provides an implementation of the FaceNet model, which is a deep convolutional neural network (CNN) designed for face recognition tasks.

Installation

1. Clone the repository:

   git clone https://github.com/Justinianus2001/face-detection.git
   

2. Go to the repository folder and set the python paths to use facenet functions

   cd face-detection
   set PYTHONPATH=[...]\face-detection\src
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Download the pre-trained detect face models for the FaceNet model and place in the model/ directory.

   Model name	LFW accuracy	Training dataset	Architecture
   20180408-102900	0.9905	CASIA-WebFace	Inception ResNet v1
   20180402-114759	0.9965	VGGFace2	Inception ResNet v1

Dataset

To train the face classification model, you will need a dataset of labeled face images. Unfortunately, this repository does not provide a pre-packaged dataset. You should prepare your own dataset with images of people's faces and their corresponding names.

Ensure that your dataset is organized in the following structure:

data/images/raw
├── person1/
│   ├── image1.jpg
│   ├── image2.jpg
│   └── ...
├── person2/
│   ├── image1.jpg
│   ├── image2.jpg
│   └── ...
└── ...

Note: Facenet model required at least 2 object detection to train and predict.

Usage

To use the face labeled system, follow these steps:

1. Prepare your input images and ensure that they are stored in a directory.

2. Run the following command to detect and recognize faces in the images:

python src/align/align_dataset_mtcnn.py data/images/raw data/images/processed --image_size 182 --margin 44 --random_order --gpu_memory_fraction 0.25

This will process the images in the data/images/raw/ directory, detect faces, and save the annotated images in the data/images/processed/ directory.

Training

To train the face classification model on your own dataset, run the following command to start the training:

python src/classifier.py TRAIN data/images/processed model/[detect-model].pb model/facemodel.pkl --batch_size 1000 --min_nrof_images_per_class 40 --nrof_train_images_per_class 35

This will start the training process using your dataset. The trained model will be saved in the specified output directory model/facemodel.pkl.

Results

Run the file src/test_webcam.py to test your result trained model online in webcam.

python src/test_webcam.py

The results achieved by the face detection and classification model trained on your dataset will vary depending on various factors such as the quality of your dataset, the size of the training set, and the chosen hyperparameters. It is recommended to experiment with different configurations to achieve the best results for your specific use case.

Contributing

Contributions to this repository are welcome. If you find any issues or would like to suggest improvements, please open an issue or submit a pull request.

License

This project is licensed under the MIT License. Feel free to use and modify the code for your own purposes.