The code files were used to achieve atmospheric correction for open ocean water from Rayleigh-corrected products.
The inputs are Rayleigh-corrected reflectance at 8 visible and 2 NIR bands, and sola, solz and relaz.
The outputs are remote sensing reflectance at 8 visible bands.
create_database.py: create training and testing data from Level-2 products
train.py: build and train the atmospheric correction model
prediction.py: predict on testing data
REQURIEMENTS: numpy h5py glob torch