Face Mask Classifier

Project Domain / Category

Image Processing/Deep Learning

Abstract / Introduction

With the ongoing concerns regarding public health and safety, particularly in the wake of the COVID-19 pandemic, ensuring compliance with mask-wearing protocols has become crucial in public and private spaces. Effective monitoring of these protocols, however, presents a significant challenge, especially in large, crowded environments. An automated solution can bridge this gap by efficiently detecting and classifying individuals based on their adherence to mask-wearing guidelines. The Face Mask Classifier capable of identifying individuals who are either wearing a mask correctly, wearing it improperly, or not wearing a mask at all. By leveraging machine learning and computer vision technologies, this classifier will provide a scalable and efficient solution for organizations seeking to enforce mask-wearing policies. The system will promote public health measures with minimal human intervention.

Functional Requirements:

Admin (Student) will perform all the following tasks.

1.      Image Input: Face Mask Classifier should support the input of images from given dataset. The system should be capable of handling various resolutions and image qualities to accommodate different imaging devices.

2.      Preprocessing: In the Face Mask Classifier project, preprocessing is essential for transforming raw input data into a suitable format for model training and inference. Preprocessing techniques should include Image Resizing, Normalization, and Handling Class Imbalance etc.

3.      Data Augmentation: To improve the robustness and generalization of the classifier, data augmentation techniques should be employed. These may include random rotations, flips, shifts, brightness adjustments, and zooming. By artificially increasing the diversity of the training dataset, the model will be better equipped to handle various real-world scenarios and face orientations.

Apply the Data Augmentation and increase the dataset by 3 folds.

4.      Model Selection and Development: Investigate different deep learning architectures and select an appropriate architecture (e.g., U-Net, 3D Convolutional Neural Networks (CNNs), V-Net, DeepMedic, Residual Networks (ResNet), DenseNet, Attention Mechanisms, YOLO ) for development.

5.      User Interface: Provide a user-friendly GUI for:

•         Uploading images.

•         Displaying results with visual indications.

•         Accessing historical data and logs.

6.      Mask Classification: The Face Mask Classifier project categorizes individuals based on their mask-wearing status to promote public health and safety. The three classification categories are:

•         With Mask: Identifies individuals wearing masks properly, covering both the nose and mouth, essential for compliance with health protocols.

•         Without Mask: Includes individuals not wearing any mask, crucial for addressing non-compliance and encouraging mask usage in public spaces.

•         Improperly Worn Mask: Captures individuals wearing masks incorrectly (e.g., below the nose or chin), enabling targeted education on proper mask usage.