Understanding and Mapping Natural Beauty

摘要: While natural beauty is often considered a subjective property of images, in this paper, we take an objective approach and provide methods for quantifying predicting the scenicness image. Using dataset containing hundreds thousands outdoor images captured throughout Great Britain with crowdsourced ratings beauty, propose to predict which explicitly accounts variance human ratings. We demonstrate that quantitative measures can benefit semantic image understanding, content-aware processing, novel application cross-view mapping, where sparsity ground-level be addressed by incorporating unlabeled overhead training prediction steps. For each application, our result qualitative improvements over baseline approaches.