DTACSNet: Onboard Cloud Detection and Atmospheric Correction With Efficient Deep Learning Models

Cesar Aybar^§, Gonzalo Mateo-García^§, Giacomo Acciarini^§, Vit Ruzicka, Gabriele Meoni, Nicolas Longepe, Luis Gómez-Chova _{^{§ development contribution}}

10.1109/JSTARS.2024.3480520

This repo contains an open implementation to run inference with DTACSNet models for atmospheric correction. This repo and trained models are released under a Creative Commons non-commercial licence

Install ⚙️:

pip install dtacs

Run:

from dtacs.model_wrapper import ACModel
model_atmospheric_correction = ACModel(model_name="CNN_corrector_phisat2")
model_atmospheric_correction.load_weights()

ac_output = model_atmospheric_correction.predict(l1c_toa_s2)

The figure above shows a sample of Sentinel-2 level 1C, DTACSNet model output and Sentinel-2 level 2A in the RGB (first row) and in the SWIR, NIR, Red (last row) composites.

See the inference tutorial for a complete example.

Citation

If you find this work useful for your research, please consider citing our work:

@article{aybar_onboard_2024,
	title = {Onboard {Cloud} {Detection} and {Atmospheric} {Correction} {With} {Efficient} {Deep} {Learning} {Models}},
	volume = {17},
	issn = {2151-1535},
	url = {https://ieeexplore.ieee.org/abstract/document/10716772},
	doi = {10.1109/JSTARS.2024.3480520},
	urldate = {2024-11-12},
	journal = {IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
	author = {Aybar, Cesar and Mateo-García, Gonzalo and Acciarini, Giacomo and Růžička, Vít and Meoni, Gabriele and Longépé, Nicolas and Gómez-Chova, Luis},
	year = {2024},
	note = {Conference Name: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
	pages = {19518--19529}
}

Acknowledgments

DTACSNet has been developed by Trillium Technologies. It has been funded by ESA Cognitive Cloud Computing in Space initiative project number D-TACS I-2022-00380.

More Cloud Detection Viz

Thick cloud Thin cloud Cloud shadow

More Atmospheric Correction Viz