*2.2. Dataset*

Satellite remote sensing data has solid spatial coverage. In remote sensing science, researchers invert images to obtain AOD and haze concentrations.

This paper hopes to simplify the data processing and replace the traditional AOD inversion method using the fitting ability of convolutional neural networks, and construct an end-to-end haze level prediction network. Considering the experimental needs, we collect several highly relevant experimental data:

