[3E4-OS-12b] 画像とAI（MIRU2019プレビュー）(2)

[3E4-OS-12b-02] Between-class Learning for Image Classification

〇Yuji Tokozume¹, Yoshitaka Ushiku¹, Tatsuya Harada^1,2 (1. The University of Tokyo, 2. RIKEN)

Keywords:Supervised Learning

We introduce our paper "Between-class Learning for Image Classification" presented at CVPR and MIRU last year. In this paper, we propose a novel learning method for image classification called between-class learning (BC learning). We generate between-class images by mixing two images belonging to different classes with a random ratio. We then input the mixed image to the model and train the model to output the mixing ratio. BC learning has the ability to impose constraints on the shape of the feature distributions, and thus the generalization ability is improved. As a result, we achieved 19.4% and 2.26% top-1 errors on ImageNet-1K and CIFAR-10, respectively.