Deep Learning for Automated Air Quality Classification Using Image Data: A Comparative Study of GoogleNet and MobileNet

The following study contributes to the improvement of environmental monitoring by providing an approach for the estimation of air quality in real-time based on deep learning models GoogleNet and MobileNet. With either compromised realtime processing capability or spatial resolution in conventional approaches, these two models can serve as practical and scalable alternatives. Empirical results demonstrate that both GoogleNet and MobileNet can classify AQI levels effectively and transfer learning significantly improves the effectiveness and generalization of comparison for tiny datasets. Class labels include “AQI-Good” to “AQI-Hazardous/Severe” Pollution information collected at sites in cities of India and Nepal along with location, time, pollutant concentration. A detailed data collection enhances the capability of models to generate accurate predictions over a range of scenarios and furthers our knowledge of temporal and spatial variability of air quality. Evaluation criteria like recall, accuracy, precision, and F1score portray the stability and reliability of the method. These measures assure their steady performance in a variety of contexts and thus guaranteeing their applicability in real life. Hence, GoogleNet and MobileNet are a few significant developments in the real-time air quality monitoring process, which provides information on would be crucial in dealing with the problems associated with air pollution.