Image classification with category centers in class imbalance situation
Yulu Zhang, Liguo Shuai, Yali Ren, Huilin Chen · 2018
In recent years, deep convolutional networks have made a milestone progress in the field of image recognition. However, the recognition ability of the deep convolution network declines in the case of unbalanced training data, in terms of categories with fewer training images, the recognition ability of the deep convolution network declines more seriously. Aiming at this kind of problem, this paper presents a new classification method which recognizes a query image by comparing distances between category centers of CNN features of the whole training dataset and the corresponding CNN feature of this query image. The experimental results on Cifar-10 and Cifar-100 show that the claimed method can more accurately identify images whose categories have only a few training samples and that the mean precision of the recognition can be improved effectively.