A Literature Study on Deep Learning Approaches for Face Mask Detection in Air Pollution Mitigation

Abstract

Air pollution ranks among the most significant environmental health issues, especially in cities with elevated pollution levels. Face masks act as an easy but effective way to limit the intake of harmful pollutants like particulate matter (PM2.5), carbon monoxide, and various atmospheric substances. Nonetheless, the efficacy of maks is significantly infuluced by regular use among the public. In this scenario, progeress in deep learning and computer vision creates a change to create automated systems for detecting face maks. This study analyzes current literature on deep learning techniques for mask detection, emphasizing their significance in minimizing air pollution. The method uses a systematic literature review through databases such as Scopus, IEEE Xplore, MDPI, and ScienceDirect, encompassing works from 2020 to 2025. Articles ware examined according to architecture, datasets, and perfomace outcomes. Results indicate that YOLOv3 and MobileNetV2 models based on CNN achieve high precision (95-99%) in detecting masks, although the majority of studies concentrate on COVID-19 scenarios. Research gaps consist of the lack of specific datasets for pollution, restricted outdoor evaluations, absence of lightweight models for edge devices, and poor integration with air quality monitoring. This research emphasizes innovation by transitioning from pandemic sotuations to enduring air pollution, palceing Al-based mask detection as a viable public health approach.