CLASS: Contemplative landscape automated scoring system

This paper presents an interdisciplinary study joining insights of landscape architecture and computer vision. In this work we used a dataset of contemplative landscape images that was collected and evaluated by experts in landscape architecture. We used the dataset to develop nine k-means clustering and one K-nearest neighbors models that are able to score landscape images based on seven different landscape image features (layers, landform, vegetation, color and light, compatibility, archetypal elements, character of peace and silence) that were identified as contributing to the overall contemplativeness of a landscape. Finally, we chose the combination of models that would produce the best combined contemplativeness score and created CLASS a scoring system that can evaluate the contemplativeness of landscape images with scores similar to those of experts.