Microplastics Dataset
收藏arXiv2025-09-30 收录
下载链接:
https://github.com/DanielPlatnick/GANsemble
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了210个聚合物光谱样本,分为10个类别,其中包括8类塑料光谱和2类非塑料光谱,这些样本是从荷兰各地水源收集而来的。为了确保数据的公正性,该数据集被划分为一个包含20个样本的未偏置测试集,以及一个包含剩余190个样本的训练集。此外,这些数据还通过极坐标转换过程被转换成了图像。在规模上,该数据集共有210个样本。任务上,我们使用深度学习技术进行数据增强和分类。
This dataset comprises 210 polymer spectral samples, categorized into 10 classes, including 8 classes of plastic spectra and 2 classes of non-plastic spectra. All samples were collected from water sources across the Netherlands. To guarantee data impartiality, the dataset is split into an unbiased test set containing 20 samples and a training set with the remaining 190 samples. Furthermore, these data were converted into images through a polar coordinate transformation process. In total, the dataset contains 210 samples. For the corresponding task, deep learning techniques are applied to implement data augmentation and classification.
搜集汇总
数据集介绍

背景与挑战
背景概述
该数据集涉及'GANsemble'项目,这是一个基于条件生成对抗网络(cGANs)的框架,专门用于生成合成微塑料数据,以解决微塑料研究领域中小型和不平衡数据集的挑战。它首次将生成式人工智能应用于微塑料数据合成,提供了基准指标如Fréchet Inception Distance(FID)和Inception Scores(IS),并包括数据增强策略选择功能。
以上内容由遇见数据集搜集并总结生成



