Mask-Wearing Detection with Computer Vision Techniques

A project leveraging computer vision and machine learning techniques to detect mask-wearing from images. This project demonstrates strong skills in image processing, feature extraction (HOG, SIFT), and classification model development (CNN, SVM, and MKP).

Project Overview

This project was part of the Master’s in Data Science program at City, University of London (2024), where it was awarded a Distinction.

Key Components

• Image Preprocessing: Techniques to enhance and standardize images for analysis.
• Feature Extraction: Utilized methods like HOG (Histogram of Oriented Gradients) and SIFT (Scale-Invariant Feature Transform) for image feature extraction.
• Model Development: Trained and evaluated machine learning models including:
  • Convolutional Neural Networks (CNNs)
  • Support Vector Machines (SVMs)
  • Multilayer Perceptrons (MLPs)
• Evaluation and Insights: Assessed models on accuracy, precision, and recall to identify the most effective approach.

Goals

• Automated Mask Detection: Develop a robust solution to identify mask-wearing from images.
• Demonstrate Proficiency: Showcase skills in computer vision and machine learning techniques.
• Provide Actionable Results: Evaluate the feasibility of mask detection for real-world applications.

Key Findings

• CNN Performance: Achieved the highest accuracy among models due to its ability to capture spatial hierarchies in images.
• Feature-Based Models: SVM and MLP models with extracted features also demonstrated good performance but required extensive preprocessing.
• Insights: High-quality image preprocessing and feature extraction significantly improved model outcomes.

Licensing

The project's code is licensed under the MIT License.

Data Availability

The datasets used in this project are not publicly available due to privacy and copyright restrictions.

• Image Datasets: Two sets of image data (training and testing) were used for model development. These datasets are excluded from this repository.
• Video Dataset: A video from the Taiwan Centres for Disease Control and Prevention was used for model testing. The video is also not included in this repository but can be accessed via the Taiwan CDC website.

For more detailed information about the datasets, refer to the data folder in this repository.