Object Cluttered Indoor Dataset (OCID)
收藏资源简介:
Object Cluttered Indoor Dataset (OCID) 是由维也纳工业大学自动化与控制研究所机器人视觉实验室创建的数据集,专注于室内环境中物体的密集杂乱场景。该数据集包含2346个场景,总计10,240个单独的物体掩码,涵盖了多种物体、背景、光照条件和视角。创建过程中,利用EasyLabel工具进行半自动像素级标注,无需预先的物体模型知识。OCID数据集主要用于评估和比较现有的物体分割方法,特别是在机器人视觉领域的实际应用中,帮助理解和解决机器人在真实世界中面临的挑战。
Object Cluttered Indoor Dataset (OCID) was developed by the Robot Vision Laboratory of the Institute of Automation and Control, Vienna University of Technology, focusing on dense cluttered object scenes in indoor environments. This dataset contains 2346 scenes and a total of 10,240 individual object masks, covering diverse objects, backgrounds, lighting conditions and viewpoints. During its construction, the EasyLabel tool was used for semi-automatic pixel-level annotation without requiring prior knowledge of object models. The OCID dataset is mainly employed to evaluate and compare existing object segmentation methods, especially in practical robotic vision applications, to help understand and resolve the challenges faced by robots in real-world scenarios.

- 1EasyLabel: A Semi-Automatic Pixel-wise Object Annotation Tool for Creating Robotic RGB-D Datasets维也纳工业大学自动化与控制研究所机器人视觉实验室 · 2019年



