Welcome to the home of the PURDUE RVL SPEC-DB color image database
Welcome to the RVL SPEC-DB color image database where you can find real color images with specular highlights along
with human-supplied groundtruth. This database
was created by Jae Byung Park in the Robot Vision Laboratory (RVL) at Purdue University. So far, it
contains over 300 color images corresponding to 100 shiny and dull objects. Some of the objects were photographed
against complex backgrounds.
Images were taken under three different illumination conditions: 1) diffuse,
2) ambient, and 3) directed lighting.
There were no restrictions on the kinds of materials the objects are made of.
We believe that the images of the RVL SPEC-DB provide sufficient diversity with regard to surface
reflectance properties. Please re-visit the SPEC-DB webpage later for updates:
http://engineering.purdue.edu/RVL/Database/specularity_database/index.html Bookmark this web page!
Easy To Download
This image database is publicly available and can be obtained from this web site. It is
free for use by academic and other not-for-profit organizations. Other individuals and/or organizations wishing
to use this database should contact RVL Webmaster. Any unauthorized
distribution of this database is prohibited.
All publications that use this database must reference the following report:
Jae B. Park and Avinash C. Kak, "A Truncated Least Squares Approach to the Detection of Specular
Highlights in Color Images", Proceedings of the IEEE International Conference on Robotics and
Automation (ICRA 2003), Taipei, Taiwan, 2003.
About the Groundtruth
At this time, the ground-truth is available for 200 of the 300 images in the database. We have used this
ground-truth for the evaluation of
specularity detection algorithms. The ground-truth at this time is based on the subjective assessments provided
by a single human. We hope to incorporate the judgements of multiple humans in the ground-truth at some
point in the future.
You are more than welcome to contact us with any comments or questions.
Good luck on your research!