Pretrained Models of the Benchmarking Algorithms for UVCGAN

来自四种图像翻译算法的预训练模型: ACL-GAN，委员会-GAN，CycleGAN和U-GAT-IT在三个基准数据集上: Selfie2Anime，CelebA_gender，CelebA_glasses。
我们训练了这些模型，以为我们在论文 “UVCGAN: UNet视觉变压器周期一致的GAN，用于不成对的图像到图像翻译” 中详细介绍的算法提供基准。
仅当基准测试算法提供预训练模型时，我们才训练模型。

Pre-trained models from four image translation algorithms, namely ACL-GAN, Committee-GAN, CycleGAN, and U-GAT-IT, were employed on three benchmark datasets: Selfie2Anime, CelebA_gender, and CelebA_glasses. We trained these models to establish baselines for the algorithm detailed in our paper titled "UVCGAN: UNet Vision Transformer-based Cycle-Consistent GAN for Unpaired Image-to-Image Translation". We only trained these models when the corresponding baseline algorithms provided their pre-trained models.

搜集汇总

背景与挑战

背景概述

该数据集提供了四种图像翻译算法（ACL-GAN、委员会-GAN、CycleGAN、U-GAT-IT）在Selfie2Anime、CelebA_gender和CelebA_glasses三个基准数据集上的预训练模型。这些模型用于支持论文'UVCGAN: UNet Vision Transformer cycle-consistent GAN for unpaired image-to-image translation'中的算法基准测试，仅在算法提供预训练模型时进行训练。