Deep Learning model to identify smoke from satellite images
Example training and testing scripts are included in the Training_Scripts and
Testing_Scripts directories.
From within the repository:
python3 -m Training_Scripts.train <variant> <suffix>
Where <variant> is one of SC, CS, S, C for spatial-channel,
channel-spatial, spatial and channel attention respectively. The <suffix> is
any string to distinguish the model save file from other files. This will create
a model file called model_<variant>_<suffix>.pt, a training log file and a
CSV file of the training and validation loss with epochs.
From within the repository:
python3 -m Testing_Scripts.evaluate <filename>
<filename> is the saved model file's name. The model variant will be taken
from the file.It will print out Cohen's Kappa, F-1 score (macro averaged),
accuracy score and the confusion matrix.