ImageNet100, Stanford Dogs, Generated-cats
收藏资源简介:
本文使用了三个数据集来验证Yuan框架的性能:ImageNet100、Stanford Dogs和自定义生成的Generated-cats数据集。ImageNet100是ImageNet数据集的子集,包含100个类别的图像;Stanford Dogs数据集专注于狗类图像,包含120个品种的狗;Generated-cats数据集则是为本文研究自定义生成的猫类图像数据集。这些数据集涵盖了广泛的视觉内容,用于测试Yuan框架在去除视觉缺陷方面的效果。数据集的大小和Tokens数未明确提及,但实验结果表明,Yuan在这些数据集上表现出色,能够有效提升生成图像的质量。数据集的应用领域主要集中在生成图像的视觉缺陷修复,旨在解决生成图像中的解剖学不准确、物体放置不当等问题,提升生成图像的美观性和实用性。
Three datasets were utilized in this study to evaluate the performance of the Yuan framework: ImageNet100, Stanford Dogs, and the custom-generated Generated-cats dataset. ImageNet100 is a subset of the ImageNet dataset, comprising images from 100 distinct categories; the Stanford Dogs dataset centers on canine imagery, encompassing 120 dog breeds; the Generated-cats dataset is a custom-built cat image dataset developed specifically for the research in this paper. These datasets span a broad spectrum of visual content and are employed to assess the effectiveness of the Yuan framework in eliminating visual defects. The scale of the datasets and the total number of Tokens are not explicitly stated, yet experimental results indicate that Yuan delivers outstanding performance across these datasets, effectively enhancing the quality of generated images. The application scope of these datasets primarily focuses on visual defect restoration for generated images, targeting issues such as anatomical inaccuracies and improper object placement in generated outputs, with the goal of improving both the aesthetic appeal and practical usability of the generated images.
数据集概述
数据集名称
AAAI 2025 - Yuan: Yielding Unblemished Aesthetics through A Unified Network for Visual Imperfections Removal in Generated Images
数据集简介
该数据集专注于通过统一的网络模型(Yuan)来去除生成图像中的视觉瑕疵,旨在提升生成图像的美学质量。数据集的具体内容和应用场景未在README文件中详细描述。
数据集来源
数据集详情页面地址:https://github.com/YuZhenyuLindy/Yuan
数据集用途
该数据集主要用于研究和开发图像生成领域的视觉瑕疵去除技术,适用于计算机视觉和图像处理相关的研究项目。
数据集特点
- 专注于生成图像中的视觉瑕疵去除。
- 使用统一的网络模型(Yuan)进行处理。
- 旨在提升生成图像的美学质量。
数据集限制
README文件中未提供数据集的具体规模、格式、使用限制等信息。

- 1Yuan: Yielding Unblemished Aesthetics Through A Unified Network for Visual Imperfections Removal in Generated Images马来亚大学计算机科学与信息技术学院 · 2025年



