Interactive Visualization and Representation Analysis Applied to Glacier Segmentation
This work addresses interpretability in earth observation for researchers and practitioners, but it is incremental as it applies existing visualization methods to a specific domain.
The researchers tackled the problem of interpreting glacier segmentation models by applying interactive visualization and representation analysis to visualize U-Net activations and build an online interface for error analysis, enabling users to discover model failure modes and providing sanity checks during data preprocessing and model training.
Interpretability has attracted increasing attention in earth observation problems. We apply interactive visualization and representation analysis to guide interpretation of glacier segmentation models. We visualize the activations from a U-Net to understand and evaluate the model performance. We build an online interface using the Shiny R package to provide comprehensive error analysis of the predictions. Users can interact with the panels and discover model failure modes. Further, we discuss how visualization can provide sanity checks during data preprocessing and model training.