PerSeg

该数据集名为PerSeg，包含了40个不同类别的对象，如日用品、动物和建筑物等。每个对象关联5至7张图像，这些图像都带有用于个性化分割评估的遮罩标注。此外，该数据集专为评估个性化分割方法而设计，包含了在不同姿态或场景下的图像。规模上，共有40个对象，每个对象都附带多张图像。所涉及的任务是个性化分割。

This dataset is named PerSeg, which encompasses 40 distinct object categories including daily necessities, animals, buildings and so forth. Each object is associated with 5 to 7 images, all with mask annotations for personalized segmentation evaluation. Moreover, this dataset is specifically designed for evaluating personalized segmentation methods, and includes images captured under varying poses or scenes. In terms of scale, there are 40 objects in total, each accompanied by multiple images. The core task involved in this dataset is personalized segmentation.

搜集汇总

背景与挑战

背景概述

PerSeg是一个为个性化分割任务设计的数据集，包含标注和图像数据，用于支持Personalize-SAM模型的训练和评估。该数据集可从云存储平台下载，并用于实现无需训练或快速微调的个性化分割方法。