CNN based Repeated Cropping for Photo Composition Enhancement

Abstract

This paper proposes a novel method for aesthetic photo recomposition using a convolutional neural network (CNN).CNN has been showing remarkable performances in various tasks, such as object detection and recognition, and we exploit its usage for photo recomposition. In our framework, CNN is used to iteratively predict cropping directions for a given photo, generating an aesthetically enhanced photo in terms of composition. Experimental results and user study show that the proposed framework can automatically crop a photo to follow specific composition guidelines, such as the rule of thirds and the salient object size.