DTACS trained models

This repository contains the trained models of the publication:

Mateo-García, G., Aybar, C., Acciarini, G., Růžička, V., Meoni, G., Longépé, N., and Gómez-Chova, L. (2023). Onboard Cloud Detection and Atmospheric Correction with Deep Learning Emulators. IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, 1875–1878. DOI: 10.1109/IGARSS52108.2023.10282605

For surface reflectance (SR) estimation we include the models:

DTACS S2 bands SR model with all the S2 bands. DTACS_SR_sentinel2.pt
DTACS phi-sat II bands SR model with overlapping bands of Phi-Sat II and Sentinel-2. DTACS_SR_phisat2.pt
DTACS Proba-V bands SR model with Blue, Red, NIR and SWIR bands. DTACS_SR_probav.pt
DTACS PlanetScope bands SR model with Blue, Green, Red and NIR bands. DTACS_SR_planetscope.pt

For cloud detection (CD):

DTACS S2 bands: CD model with all the S2 bands. DTACS_CLOUD_ALL.pt
DTACS RGBNIR bands: CD model with Red, Green, Blue and NIR bands. DTACS_CLOUD_RGBNIR.pt

Examples of use here: https://github.com/spaceml-org/DTACSNet

Licence

All pre-trained models in this repository are released under a Creative Commons non-commercial licence