# Attentive Normalization for Conditional Image Generation

arXiv: https://arxiv.org/abs/2004.03828 · HTML: https://mlsanity.com/paper/2004.03828/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Apr 2020.

## Built on by (most cited first, 10 of 10)

- [Image Synthesis via Semantic Composition](https://mlsanity.com/paper/2109.07053/) (2021)
- [Improving Visual Quality of Image Synthesis by A Token-based Generator with Transformers](https://mlsanity.com/paper/2111.03481/) (2021)
- [A^2-FPN: Attention Aggregation based Feature Pyramid Network for Instance Segmentation](https://mlsanity.com/paper/2105.03186/) (2021)
- [PalGAN: Image Colorization with Palette Generative Adversarial Networks](https://mlsanity.com/paper/2210.11204/) (2022)
- [What Makes for Good Tokenizers in Vision Transformer?](https://mlsanity.com/paper/2212.11115/) (2022)
- [Sandwich Batch Normalization: A Drop-In Replacement for Feature Distribution Heterogeneity](https://mlsanity.com/paper/2102.11382/) (2021)
- [Can3Tok: Canonical 3D Tokenization and Latent Modeling of Scene-Level 3D Gaussians](https://mlsanity.com/paper/2508.01464/) (2025)
- [Self-Supervised Domain Adaptation for Diabetic Retinopathy Grading using Vessel Image Reconstruction](https://mlsanity.com/paper/2107.09372/) (2021)
- [TESL-Net: A Transformer-Enhanced CNN for Accurate Skin Lesion Segmentation](https://mlsanity.com/paper/2408.09687/) (2024)
- [IR-GAN: Image Manipulation with Linguistic Instruction by Increment Reasoning](https://mlsanity.com/paper/2204.00792/) (2022)

## Abstract (arXiv)

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.