ARTeFACT
收藏资源简介:
ARTeFACT数据集由格拉斯哥大学创建,专注于模拟媒体损坏检测,涵盖15种不同类型的损坏,涉及多种主题、媒体和历史背景。数据集包含418张高分辨率图像,每张图像都有像素级的损坏标注,总共有超过11,000个标注。数据集的创建过程包括专家的手动标注和多轮审核,确保标注质量。该数据集主要应用于文化遗产保护领域,旨在解决模拟媒体损坏检测和修复的问题,为评估和改进现有及未来的分割模型提供了全面的基准。
The ARTeFACT dataset, created by the University of Glasgow, focuses on analog media damage detection, covering 15 distinct types of damage across diverse themes, media formats, and historical contexts. It contains 418 high-resolution images, each paired with pixel-level damage annotations, totaling over 11,000 annotations. The dataset was developed through expert manual annotation and multi-round reviews to ensure annotation quality. Primarily applied in the field of cultural heritage conservation, this dataset aims to address the challenges of analog media damage detection and restoration, providing a comprehensive benchmark for evaluating and improving both existing and future segmentation models.

- 1ARTeFACT: Benchmarking Segmentation Models on Diverse Analogue Media Damage格拉斯哥大学 · 2024年



