Face-Identification-using-CNNs

This repository comprises a face identification(multi-class classification) using a subset of the Labeled Faces in the Wild (LFW) dataset.

Overview

This repository explores different approaches to face identification using a subset of the [Labeled Faces in the Wild (LFW) dataset]. The primary focus is on comparing the performance of traditional Eigenfaces with modern Convolutional Neural Network (CNN) models.

Objectives

• Evaluate the effectiveness of Eigenfaces, a Principal Component Analysis (PCA)-based method, for face identification.
• Compare Eigenfaces with state-of-the-art CNN models in terms of accuracy and robustness.
• Investigate the impact of varying CNN architectures, hyperparameters, and optimization techniques on face identification.

Eigenfaces Overview

Eigenfaces is a Principal Component Analysis (PCA)--based dimensionality reduction technique used for face identification. It extracts features from facial images and reduces the dimensionality of the dataset.

Logistic Regression and Eigenfaces

Eigenfaces and Logistic Regression is a traditional method for face identification. It involves flattening images and applying Principal Component Analysis (PCA) to reduce dimensionality, followed by logistic regression for classification.

• Architecture

CNN Models

CNN Model 1

• Architecture:

  

• Hyperparameters:

  • Optimizer: adam
  • Loss: Sparse categorical cross-entropy
  • Batch Size: 64
  • Epoch: 15

CNN Model 2

• Architecture:

  

• Hyperparameters:

  • Optimizer: RMSProp
  • Loss: Sparse categorical cross-entropy
  • Batch Size: 128
  • Epoch: 20

CNN Model 3

• Architecture:

  

• Hyperparameters:

  • Optimizer: Adam
  • Loss: Sparse categorical cross-entropy
  • Batch Size: 64
  • Epoch: 10

CNN Model 4

• Architecture:

  

• Hyperparameters:

  • Optimizer: adam
  • Loss: Sparse categorical cross-entropy
  • Batch Size: 64
  • Epoch: 20

Getting Started

# Clone the repository
git clone https://github.com/your-username/drowsiness-detection.git
cd face-mask-detection

# Install the required dependencies
pip install -r requirements.txt

Results

The notebook provides visualizations of correctly and incorrectly identified images along with their corresponding training images.

Correctly Identified Image

Incorrectly Identified Image

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Acknowledgments

• Dataset: Labeled Faces in the Wild (https://vis-www.cs.umass.edu/lfw/)