Fig. 4From: Refining deep convolutional features for improving fine-grained image recognitionComparison of distribution of data on VLAD, Fisher Vectors where c, k denote number of channels, number of centers of K-means, or GMM, respectivelyBack to article page