B3DO: Berkeley 3-D Object Dataset

A quality depth sensor, the Microsoft Kinect, is now in millions of homes. Yet robust household object detection is still not a reality. To get there, we are collecting a massive, crowd-sourced, and challenging 3-D object dataset.

Download the dataset

Version 1 of the dataset is available as a zip archive here. Results that refer to this version are available in the following publication:
A. Janoch, S. Karayev, Y. Jia, J.T. Barron, M. Fritz, K. Saenko, T. Darrell. A Category-Level 3-D Object Dataset: Putting the Kinect to Work. ICCV Workshop on Consumer Depth Cameras in Computer Vision 2011.

Details

For the first few decades of the field's existence, computer vision has been focused on algorithmic, logical approaches to perception. But it was only with the arrival of the era of large datasets, collected and annotated using the Internet, that truly significant progress has been made in object recognition.

With the recent proliferation of the Microsoft Kinect, we can for the very first time collect an "Internet-sized" dataset of matched color and depth images. We are making this dataset public to hasten progress in the field of robust, cheap, household object recognition.

Current statistics of object distribution in our dataset.

Example of a color, registered, and smoothed depth images in our dataset.

Thank you for your interest.

Made by Sergey Karayev & Allison Janoch.