OCID Object Segmentation Dataset

OCID stands for Object Clutter Indoor Dataset. It is an RGB-D dataset that contains point clouds with point-by-point annotations for each object. These data were taken with two ASUS-PRO Xtion cameras at different heights. The dataset captures different settings of objects, backgrounds, environments, sensor-scene distances, viewing angles, and lighting conditions.

OCID aims to systematically compare existing object segmentation methods in scenes with increasing clutter. In addition, OCID also provides real data for other vision tasks such as object classification and recognition.