INPUTS

This dataset used for this model is [nadh0708/]ECMWF-processed-oxygen-data](nadh0708/ECMWF-processed-oxygen-data) which consist of input tiles of 224x224x6 (224x224 image with 6 bands). The bands are:

Sea surface height above geoid (zos [m])
Mole concentration of dissolved molecular oxygen in sea watero2 [mmol/m3]
Sea water salinity (so [10-3])
Sea water potential temperature (thetao [°C])
Eastward sea water velocity (uo [m/s])
Northward sea water velocity (vo [m/s])

MODELS

This is a fine-tuned Prithvi-100 model. There are five models in this repository:

MAE_nonScheduler or pred_MAE : L1Loss + no training the encoding blocks of base model
MAE_nonScheduler_differentLR or MAE_finetune_partial : L1Loss + training the encoding blocks of base model
MSE_nonScheduler or pred_MSE : MSELoss + no training the encoding blocks of base model
MSE_nonScheduler_differentLR or MSE_finetune_partial : MSELoss + training the encoding blocks of base model

INFERENCE

Those models can be used with this google colab repository https://colab.research.google.com/drive/19_tv46nW4_9GbM0QNID80Zo4I43fu2OE

CITATION

@article{Prithvi-100M-preprint, author = {Jakubik, Johannes and Roy, Sujit and Phillips, C. E. and Fraccaro, Paolo and Godwin, Denys and Zadrozny, Bianca and Szwarcman, Daniela and Gomes, Carlos and Nyirjesy, Gabby and Edwards, Blair and Kimura, Daiki and Simumba, Naomi and Chu, Linsong and Mukkavilli, S. Karthik and Lambhate, Devyani and Das, Kamal and Bangalore, Ranjini and Oliveira, Dario and Muszynski, Michal and Ankur, Kumar and Ramasubramanian, Muthukumaran and Gurung, Iksha and Khallaghi, Sam and Li, Hanxi (Steve) and Cecil, Michael and Ahmadi, Maryam and Kordi, Fatemeh and Alemohammad, Hamed and Maskey, Manil and Ganti, Raghu and Weldemariam, Kommy and Ramachandran, Rahul}, month = oct, title = {{Foundation Models for Generalist Geospatial Artificial Intelligence}}, journal = {Preprint Available on arxiv:2310.18660}, year = {2023} }