具体来说,生成器使用了一个混合的UNet-ViT结构,判别器使用了一个特殊的头部(batch head)来增强判别能力,训练过程使用了自监督预训练和对比损失等技术。 显著地超越了UVCGAN和EGSDE:指的是在一些公开的无配对图像到图像转换的数据集上,如CelebA和CelebA-HQ,UVCGAN v2的生成图像的真实性和多样性都显著地优于UVCGAN...
UVCGAN v2: An Improved Cycle-Consistent GAN for Unpaired Image-to-Image Translation - LS4GAN/uvcgan2
uvcgan data/data: Add CycleGANv2 dataset selector Jun 20, 2023 .gitignore Public release of UVCGAN code Mar 10, 2022 .pylintrc Public release of UVCGAN code Mar 10, 2022 CITATIONS CITATIONS: Add citations file Feb 3, 2023 LICENSE LICENSE: Specify license applicability limits Dec 31, 202...
), you may find it more useful to take one of the UVCGAN or UVCGANv2 training scripts as a starting point. To make those scripts work with your dataset, simply modify the path parameter of the data configuration. The path should point to the location of your dataset on a disk. 0.2 ...