Automatic detection of floating aquatic vegetation from remote sensing data

Install deps:

poetry install

Train Barlow Twins

python -m train.barlowtwins.train -h

To run the training with EuroSAT you need to Download EuroSAT:

Download link

Then you generate a .csv with the train split:

python -m dataset.EuroSAT.generate_train_test_split /path/to/eurosat/folders

Feed the train.csv path to the train script

Train linear classifier with EuroSAT

Follow the jupyter notebook finetune_eurosat-linear.ipynb to replicate it locally with EuroSAT.

Train linear classifier with aquatic vegetation dataset

First the data needs to be downloaded runnning:

python -m scripts.download

You have to confiure the main function to fit your local system and also need an account for DHub from the Copernicus program.

Once you have the dataset downloaded you can run the finetunning process.

Follow the jupyter notebook finetune_mat-linear.ipynb to replicate it locally with the local dataset.

Visualize dataset

Also in the inference folder couple of jupyter notebooks can be found that allows you to use the model to do image serach in a tile.