A Category-Level 3-D Object Dataset: Putting the Kinect to Work
Title | A Category-Level 3-D Object Dataset: Putting the Kinect to Work |
Publication Type | Conference Paper |
Year of Publication | 2011 |
Authors | Janoch, A., Karayev S., Jia Y., Barron J. T., Fritz M., Saenko K., & Darrell T. |
Page(s) | 1168-1174 |
Other Numbers | 3234 |
Abstract | Recent proliferation of a cheap but quality depth sensor,the Microsoft Kinect, has brought the need for a challengingcategory-level 3D object detection dataset to thefore. We review current 3D datasets and find them lackingin variation of scenes, categories, instances, and viewpoints.Here we present our dataset of color and depthimage pairs, gathered in real domestic and office environments.It currently includes over 50 classes, with moreimages added continuously by a crowd-sourced collectioneffort. We establish baseline performance in a PASCALVOC-style detection task, and suggest two ways that inferredworld size of the object may be used to improve detection.The dataset and annotations can be downloaded athttp://www.kinectdata.com. |
URL | http://www.icsi.berkeley.edu/pubs/vision/categorylevel11.pdf |
Bibliographic Notes | Proceedings of the First IEEE Workshop on Consumer Depth Cameras for Computer Vision at the International Conference on Computer Vision (ICCV 2011), pp. 1168-1174, Barcelona, Spain |
Abbreviated Authors | A. Janoch, S. Karayev, Y. Jia, J. Y. Barron, M. Fritz, K. Saenko, and T. Darrell |
ICSI Research Group | Vision |
ICSI Publication Type | Article in conference proceedings |