Constraint-Based Evaluation of Map Images Generalized by Deep Learning

Deep learning techniques have recently been experimented for map generalization. Although promising, these experiments raise new problems regarding the evaluation of output images. Traditional generalization cannot directly be applied to results in a raster format. Additionally, internal used by deep models is mostly based on realism images and accuracy pixels, none criteria sufficient evaluate process. Finally, processes tend hide causal mechanisms do not always guarantee result that follows cartographic principles. In this article, we propose method adapt constraint-based generated models. We focus use case mountain road generalization, detail seven raster-based constraints, namely, clutter, coalescence reduction, smoothness, position preservation, connectivity noise absence, color constraints. These constraints can contribute current studies learning-based as they help guide process, compare different models, validate identify remaining They also assess quality training examples.