LAP-GAN: Label augmentation with perceptual loss for self-supervised text-to-image synthesis

Open Webpage

Yong Xuan Tan, Jit Yan Lim, Kian-Ming Lim, Chin-Poo Lee

Published: 01 Jan 2026, Last Modified: 04 Nov 2025Expert Syst. Appl. 2026EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A novel self-supervised text-to-image synthesis approach with label augmentation.•A label augmented discriminator is introduced for goal consistency among tasks.•A perceptual loss is introduced to capture accurate features in rotated samples.•A cycleGAN architecture is introduced to improve the image and text alignment.