Data access

[gabrieltseng]

Data is currently stored in a private google cloud bucket. We will need to move it to a more generally accessible location (likely Zenodo).

• [ ] Data in Zenodo
• [ ] Download functions updated to pull from Zenodo instead of google cloud

[GISScience]

or, upload in google drive or drop box. please also provide the pre-trained model to check.

[gabrieltseng]

or, upload in google drive or drop box. please also provide the pre-trained model to check.

The pretrained model is available on github. From #4 :

The trained model is available here: https://github.com/nasaharvest/presto/blob/main/data/default_model.pt You can load the pretrained model with the following code:

from presto import Presto
encoder_decoder = Presto.load_pretrained()

where load_pretrained just loads the state dict in default_model.pt.

[Struggling10000]

Could you please provide the datasets used in the eval model？

[gabrieltseng]

Hi @Struggling10000 ,

Which datasets are you interested in? 3 of the datasets are publicly available:

1. EuroSat (the readme has instructions on downloading)
2. TreeSat, which should be downloaded into this folder (I'll add some instructions making this explicit update: instructions added in the readme).
3. CropHarvest, which is automatically downloaded when you run the CropHarvest eval task.

Adding the fuel moisture and algae blooms data to Zenodo is on my todos!