Datasets

CMU Kitchen Occlusion Dataset (CMU_KO8)

This dataset contains 1600 images of 8 texture-less household items (i.e. cup, pitcher, shaker, thermos, shaker, scissors, baking pan) under severe occlusions in cluttered, kitchen enviornments. The data is split for object instance detection from a single-view of an object and from multiple-views of an object. For the single-view portion of the dataset, the occluded regions are labeled and the occlusion statistics are provided for each image.

Download 213 MB

Reference:
E. Hsiao and M. Hebert. Occlusion reasoning for object detection under arbitrary viewpoint. In CVPR, 2012. [pdf][bibtex]

CMU Grocery Dataset (CMU10_3D)

This dataset contains 620 images of 10 grocery items (i.e. soup cans, soda cans, cartons, boxes) in natural, kitchen environments. The objects in this dataset are seen under a wide range of viewpoints, lighting condition changes and clutter. The dataset includes 500 images with ground truth 2D segmentations and 120 images with ground truth 6D pose.

Download 83.9 MB

Reference:
E. Hsiao, A. Collet and M. Hebert. Making specific features less discriminative to improve point-based 3D object recognition. In CVPR, 2010. [pdf][bibtex]