Next Article in Journal
From Design to Deployment: A Comprehensive Review of Theoretical and Experimental Studies of Multi-Energy Systems for Residential Applications
Previous Article in Journal
Ultra-Dense Uplink UAV Lossy Communications: Trajectory Optimization Based on Mean Field Game
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
This is an early access version, the complete PDF, HTML, and XML versions will be available soon.
Article

MD-GAN: Multi-Scale Diversity GAN for Large Masks Inpainting

1
Key Laboratory of Henan Province, Computer and Information Engineering College, “Educational Artificial Intelligence and Personalized Learning”, Xinxiang 453007, China
2
Information Management Department, Henan Normal University, Xinxiang 453007, China
*
Author to whom correspondence should be addressed.
Electronics 2025, 14(11), 2218; https://doi.org/10.3390/electronics14112218
Submission received: 10 April 2025 / Revised: 26 May 2025 / Accepted: 28 May 2025 / Published: 29 May 2025

Abstract

Image inpainting approaches have made considerable progress with the assistance of generative adversarial networks (GANs) recently. However, current inpainting methods are incompetent in handling the cases with large masks and they generally suffer from unreasonable structure. We find that the main reason is the lack of an effective receptive field in the inpainting network. To alleviate this issue, we propose a new two-stage inpainting model called MD-GAN, which is a multi-scale diverse GAN. We inject dense combinations of dilated convolutions in multiple scales of inpainting networks to obtain more effective receptive fields. In fact, the result of inpainting large masks is generally not uniquely deterministic. To this end, we newly propose the multi-scale probabilistic diverse module, which achieves diverse content generation by spatial-adaptive normalization. Meanwhile, the convolutional block attention module is introduced to improve the ability to extract complex features. Perceptual diversity loss is added to enhance diversity. Extensive experiments on benchmark datasets including CelebA-HQ, Places2 and Paris Street View demonstrate that our approach is able to effectively inpaint diverse and structurally reasonable images.
Keywords: image diverse inpainting; generative adversarial networks; spatial adaptive normalization; dilated convolution; large missing areas image diverse inpainting; generative adversarial networks; spatial adaptive normalization; dilated convolution; large missing areas

Share and Cite

MDPI and ACS Style

Wang, S.; Guo, X.; Guo, W. MD-GAN: Multi-Scale Diversity GAN for Large Masks Inpainting. Electronics 2025, 14, 2218. https://doi.org/10.3390/electronics14112218

AMA Style

Wang S, Guo X, Guo W. MD-GAN: Multi-Scale Diversity GAN for Large Masks Inpainting. Electronics. 2025; 14(11):2218. https://doi.org/10.3390/electronics14112218

Chicago/Turabian Style

Wang, Shibin, Xuening Guo, and Wenjie Guo. 2025. "MD-GAN: Multi-Scale Diversity GAN for Large Masks Inpainting" Electronics 14, no. 11: 2218. https://doi.org/10.3390/electronics14112218

APA Style

Wang, S., Guo, X., & Guo, W. (2025). MD-GAN: Multi-Scale Diversity GAN for Large Masks Inpainting. Electronics, 14(11), 2218. https://doi.org/10.3390/electronics14112218

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop