Class Imbalance in Satellite Mapping: Rare Targets Need Explicit Evaluation
Class imbalance occurs when some classes occupy far fewer samples or area than others.
THE ITSENSING JOURNAL
Practical satellite reviews, remote sensing methods and GIS workflows. Explore the science, understand the limits, and take your next project into iT Sensing.

Class imbalance occurs when some classes occupy far fewer samples or area than others.
A model score, vote fraction and calibrated probability are different quantities.
Semantic segmentation assigns classes across image pixels using spatial context.
Object detection predicts locations and classes of defined objects.
Regression estimates a continuous target from imagery and other predictors.
Feature importance describes how a fitted model uses predictors under a particular calculation.
A change model compares observations or derived features across time.
Counting classified pixels gives map area under the classifier’s decisions.
Model drift describes changing input or performance conditions after deployment.
GeoTIFF combines raster values with georeferencing and related metadata.
A Cloud Optimized GeoTIFF organizes imagery so compatible clients can request useful byte ranges and overview levels.
A STAC Item describes a spatiotemporal asset using a GeoJSON-based structure and links to data.
No matching articles. Try a different search.