A deep learning and image-based model for air quality estimation.

The serious threat of air pollution to human health makes air quality a focus of public attention, and effective, timely air quality monitoring is critical to pollution control and human health. This paper proposes a deep learning and image-based model for air quality estimation. The model extracts feature information from scene images captured by camera equipment and then classifies them to estimate air quality levels. A self-supervision module (SCA) is added to the model and the global context information of the feature map is used to reconstruct the features by using the interdependence between the channel maps to enhance the interdependent channel maps and improve the ability of feature representation. In addition, a high-quality outdoor air quality data set (NWNU-AQI) was compiled to facilitate the training and evaluation of the model's performance. This paper compares and analyzes AQC-Net, Support Vector Machine (SVM), and Deep Residual Network (ResNet) on NWNU-AQI. The experimental results show that AQC-Net yields more accurate results for air quality classification than other methods.