Large margin deep embedding for aesthetic image classification

Guanjun Guo, Hanzi Wang, Yan Yan, Liming Zhang, Bo Li

Abstract: We present an LMDE method with a novel network structure and an effective joint loss function, which takes advantage of both the triplet loss function and the hinge loss function. The minimization of the joint loss function ensures that the intra-class variability of the features belonging to the same class is reduced and the inter-class separability of the features from different classes is increased. As shown in the experiments, the proposed LMDE method significantly outperforms several other state-of-the-art aesthetic classification methods in terms of classification accuracy.