Model Generalization in Deep Learning Applications for Land Cover Mapping
This highlights a critical limitation for researchers and practitioners using satellite imagery for land-use classification, as it reveals that models trained on specific data may not reliably transfer to new contexts.
The study found that deep learning models for land cover mapping show high variability in performance when applied to out-of-sample continents or seasons, indicating poor generalization across different geographic and temporal conditions.
Recent work has shown that deep learning models can be used to classify land-use data from geospatial satellite imagery. We show that when these deep learning models are trained on data from specific continents/seasons, there is a high degree of variability in model performance on out-of-sample continents/seasons. This suggests that just because a model accurately predicts land-use classes in one continent or season does not mean that the model will accurately predict land-use classes in a different continent or season. We then use clustering techniques on satellite imagery from different continents to visualize the differences in landscapes that make geospatial generalization particularly difficult, and summarize our takeaways for future satellite imagery-related applications.