Attentive normalization for conditional image generation

Jan 1, 2020·
Yi Wang
Ying-Cong Chen
Ying-Cong Chen
Xiangyu Zhang
Jian Sun
Jiaya Jia
· 0 min read
Traditional convolution-based generative adversarial networks synthesize images based on hierarchical local operations, where long-range dependency relation is implicitly modeled with a Markov chain. It is still not sufficient for categories with complicated structures. In this paper, we characterize long-range dependence with attentive normalization (AN), which is an extension to traditional instance normalization. Specifically, the input feature map is softly divided into several regions based on its internal semantic similarity, which are respectively normalized. It enhances consistency between distant regions with semantic correspondence. Compared with self-attention GAN, our attentive normalization does not need to measure the correlation of all locations, and thus can be directly applied to large-size feature maps without much computational burden. Extensive experiments on class-conditional image generation and semantic inpainting verify the efficacy of our proposed module.
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (Oral, Acceptance Rate: 5.7%)