Yuru-chara image dataset
收藏arXiv2020-04-17 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2004.08066v1
下载链接
链接失效反馈官方服务：
资源简介：
Yuru-chara图像数据集是由东京农业技术大学创建的，用于研究生成对抗网络（GANs）在生成吉祥物角色图像中的应用。该数据集包含4018个样本，这些样本是从‘Yuru-chara Grand Prix’收集并经过处理的，每个图像的分辨率为144×128像素。数据集创建过程中，对原始图像进行了背景和工具的移除，以及多角色图像的分离处理。该数据集主要用于训练和评估基于GAN的吉祥物角色生成模型，旨在解决小数据集上GAN训练的稳定性和生成图像质量问题。
The Yuru-chara Image Dataset was developed by Tokyo University of Agriculture and Technology for research on the application of Generative Adversarial Networks (GANs) in generating mascot character images. This dataset comprises 4018 samples collected and processed from the 'Yuru-chara Grand Prix', with each image having a resolution of 144 × 128 pixels. During the dataset construction process, background and extraneous tools were removed from the original images, and multi-character images were separated into individual character samples. This dataset is primarily utilized for training and evaluating GAN-based mascot character generation models, with the goal of addressing the challenges of GAN training stability and generated image quality when working with small-scale datasets.
提供机构：
东京农业技术大学
创建时间：
2020-04-17
