Right now, Accuracy metric for the model is pixelwise; but with DBSCAN we can potentially cluster disjoint pionts and create a hull with them that encompasses a larger area than just pixels. This area can be compared with the area of truth data (verified shapefiles) to give out better accuracy metric.
Right now, Accuracy metric for the model is pixelwise; but with DBSCAN we can potentially cluster disjoint pionts and create a hull with them that encompasses a larger area than just pixels. This area can be compared with the area of truth data (verified shapefiles) to give out better accuracy metric.