论文阅读 [CVPR-2022] Attribute Group Editing for Reliable Few-shot Image Generation

论文阅读 [CVPR-2022] Attribute Group Editing for Reliable Few-shot Image Generation

studyai.com

搜索论文: Attribute Group Editing for Reliable Few-shot Image Generation

http://www.studyai.com/search/whole-site/?q=Attribute+Group+Editing+for+Reliable+Few-shot+Image+Generation

摘要(Abstract)

Few-shot image generation is a challenging task even using the state-of-the-art Generative Adversarial Networks (GANs).

即使使用最先进的生成性对抗网络（GAN），生成少镜头图像也是一项具有挑战性的任务。

Due to the unstable GAN training process and the limited training data, the generated images are often of low quality and low diversity.

由于GAN的训练过程不稳定，训练数据有限，生成的图像质量和多样性往往较低。

In this work, we propose a new “editing-based” method, i.e., Attribute Group Editing (AGE), for few-shot image generation.

在这项工作中，我们提出了一种新的基于编辑的方法，即属性组编辑（AGE），用于生成少量镜头的图像。

The basic assumption is that any image is a collection of attributes and the editing direction for a specific attribute is shared across all categories.

基本假设是，任何图像都是属性的集合，特定属性的编辑方向在所有类别中共享。

AGE examines the internal representation learned in GANs and identifies semantically meaningful directions.

AGE检查在GANs中学习到的内部表征，并确定语义上有意义的方向。

Specifically, the class embedding, i.e., the mean vector of the latent codes from a specific category, is used to represent the category-relevant attributes, and the category-irrelevant attributes are learned globally by Sparse Dictionary Learning on the difference between the sample embedding and the class embedding.

具体来说，类别嵌入，即来自特定类别的潜在代码的平均向量，用于表示类别相关属性，并且通过稀疏字典学习，根据样本嵌入和类别嵌入之间的差异，全局学习与类别无关的属性。

Given a GAN well trained on seen categories, diverse images of unseen categories can be synthesized through editing category irrelevant attributes while keeping category-relevant attributes unchanged.

给定一个在已知类别上受过良好训练的GAN，可以通过编辑与类别无关的属性，同时保持与类别相关的属性不变，来合成未知类别的各种图像。

Without re-training the GAN, AGE is capable of not only producing more realistic and diverse images for downstream visual applications with limited data but achieving controllable image editing with interpretable category-irrelevant directions.

无需对GAN进行重新训练，AGE不仅能够用有限的数据为下游视觉应用程序生成更真实、更多样的图像，而且能够实现可控的图像编辑，并具有可解释的类别无关方向。

Code is available at https://github.com/UniBester/AGE.