Variable Annotations from a Crowdsourcing Experiment on Segmentation (VACES) dataset

This dataset contains images with semantic segmentation masks produced by crowdsourced annotators on Amazon Mechanical Turk. There are multiple masks for each image reflecting different configurations of the crowdsourcing task. The purpose of this dataset is to promote the study of how crowdsourcing task configuration impacts variation in label quality for image segmentation.

本数据集收录了由亚马逊机械 Turk（Amazon Mechanical Turk）平台上的众包标注人员生成的带有语义分割掩码（semantic segmentation masks）的图像。每张图像对应多个掩码，这些掩码反映了众包任务的不同配置方案。本数据集旨在推动相关研究，探究众包任务配置如何影响图像分割任务中标注质量的差异。