Rethinking cross-domain semantic relation for few-shot image generation

Training well-performing Generative Adversarial Networks (GANs) with limited data has always been challenging. Existing methods either require sufficient data (over 100 training images) for training or generate images of low quality and low diversity. To solve this problem, we propose a new Cross-do...

Full description

Saved in:
Bibliographic Details
Published in:Applied intelligence (Dordrecht, Netherlands) Vol. 53; no. 19; pp. 22391 - 22404
Main Authors: Gou, Yao, Li, Min, Lv, Yilong, Zhang, Yusen, Xing, Yuhang, He, Yujie
Format: Journal Article
Language:English
Published: New York Springer US 01.10.2023
Springer Nature B.V
Subjects:
Ablation
Artificial Intelligence
Computer Science
Generative adversarial networks
Image processing
Image quality
Machines
Manufacturing
Mechanical Engineering
Processes
Semantic relations
Semantics
Training
Cross-domain semantic relation
Generative adversarial networks
Few-shot image generation
Perceptual similarity
Contrastive learning
ISSN:0924-669X, 1573-7497
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Training well-performing Generative Adversarial Networks (GANs) with limited data has always been challenging. Existing methods either require sufficient data (over 100 training images) for training or generate images of low quality and low diversity. To solve this problem, we propose a new Cross-domain Semantic Relation (CSR) loss. The CSR loss improves the performance of the generative model by maintaining the relationship between instances in the source domain and generated images. At the same time, a perceptual similarity loss and a discriminative contrastive loss are designed to further enrich the diversity of generated images and stabilize the training process of models. Experiments on nine publicly available few-shot datasets and comparisons with the current nine methods show that our approach is superior to all baseline methods. Finally, we perform ablation studies on the proposed three loss functions and prove that these three loss functions are essential for few-shot image generation tasks. Code is available at https://github.com/gouayao/CSR .
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0924-669X
1573-7497
DOI:10.1007/s10489-023-04602-8