Real-World Shelf Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://sites.google.com/view/stow-corl23
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了在实际场景中收集并手动标注的真实世界数据，涵盖了44个序列，共计220张货架场景的图片。此外，数据集中包含了超过150个不同种类的物体，物体数量会逐渐增加，直到储物箱装满或桌面上有大约10个物体为止。该数据集的规模为44个序列，220张图片。其任务是对未见过的物体进行离散帧分割和跟踪。
This dataset contains real-world data collected in actual scenarios and manually annotated, covering 44 sequences with a total of 220 shelf-scene images. Additionally, the dataset includes over 150 distinct object categories, with the number of objects increasing gradually until the storage bin is full or there are approximately 10 objects on the desktop. The scale of this dataset is 44 sequences and 220 images, and its task is discrete frame segmentation and tracking for unseen objects.
提供机构：
Authors of the paper
