Table 3 Model Validation against Kuro Siwo global dataset

From: Mapping global floods with 10 years of satellite radar data

Model

Prec

Rec

F1

IOU

AI4G Model

0.84

0.72

0.77

0.63

GFM

0.73

0.70

0.72

0.56

Kuro Siwo baseline

  

0.75–0.80

 
  1. Model performance against the test set in the Kuro Siwo dataset. Our model performance exceeds that of the Global Flood Monitoring (GFM) data and is on par with the models trained on the Kuro Siwo dataset itself.
Back to article page