OmniArt

OmniArt 是一个艺术数据集，包含超过 15 个在线艺术品收藏和上传到互联网上的用户生成的艺术。涵盖从公元前 157 年（共同时代之前）到 2017 年 8 月 28 日的时期，它包含 500 多种类型的艺术品，这些艺术品归属于全球 36000 多名艺术家。其他元数据包含但不限于风格、学校、Iconclass、颜色代码和托盘、材料、当前位置、实际尺寸、技术等。所有包含的图像都以它们起源的集合中的本机形式保存，随附的元数据保存在具有相同 ID 的 .json 文件中。对于每个数据集条目，都有一个持久的元数据文件，该文件会被清理和解析，以便更快地集成到机器学习项目中。如果数据集条目的属性未知，则将其标记为未知。

当前骨架数据集大小为 513 Gb，由 .jpg 图像和 .json 元数据文件组成。额外的 550 Gb 存档可应要求提供额外的特定于收集的元数据、提取的特征、快速 sql 沉浸式片段和训练模型。鉴于数据集的大小，我们根据请求生成下载链接。

OmniArt is an art dataset comprising over 15 online art collections and user-generated artworks uploaded to the Internet. Covering the period from 157 Before Common Era (BCE) to August 28, 2017, it includes more than 500 types of artworks created by over 36,000 artists worldwide. Additional metadata includes, but is not limited to, style, school, Iconclass, color codes and palettes, materials, current location, physical dimensions, techniques, and more. All included images are preserved in their native source collection formats, with accompanying metadata stored in .json files sharing the same ID. For each dataset entry, a persistent metadata file is cleaned and parsed to enable faster integration into machine learning projects. If an attribute of a dataset entry is unknown, it will be marked as "unknown".

The current skeleton dataset has a size of 513 GB, composed of .jpg images and .json metadata files. An additional 550 GB archive is available upon request, providing extra collection-specific metadata, extracted features, fast SQL-accessible snippets, and trained models. Given the considerable size of the dataset, download links are generated per request.

搜集汇总

背景与挑战

背景概述

OmniArt是一个大型艺术数据集，汇集了超过15个在线收藏和用户生成的艺术品，时间范围从公元前157年至2017年，涵盖500多种类型和36000多名艺术家，并提供风格、颜色等丰富元数据。数据集由图像和JSON文件构成，核心大小为513GB，还可根据请求提供额外的550GB存档数据。