Download our dataset and put
your algorithms to the test.
The Classification Blind Test project aims to provide a standard hyperspectral dataset to the remote sensing community. Additionally, the project will score the ability of classification algorithms used on the dataset provided. These services are provided for free — after registering, anyone can access this project's data and evaluation services.
The downloadable dataset includes the "test" image, its header file, and a small rectangle of true "training" RoIs (Regions of Interest). Once the algorithm produces a result, the class image can be uploaded to this site and will be scored against truth data which we will keep private. This happens in real time, and can be compared with other algorithms used on the data.
Up until now, there have been few examples where hyperspectral information has been provided in the free and easily accessible way that the Target Detection Blind Test, and this Classification Blind Test, provide. For people who wish to simply create classification algorithms and don't want to worry about going out to the field and collecting data, this is a perfect fit — the data has already been collected and processed, all you need to do is download it. After your download, we encourage you to test your classification algorithms on the supplied data. To reduce the potential for biased results, we have withheld the truth information and included a small area of training information. This data contains no restricted information and the rights to it are possessed by the organizations that fund this project. Because of this, it can be used by anyone who wishes to use it, and it may be used in publications.
This project is based in the Digital Imaging and Remote Sensing laboratory, a research group that is a part of the Chester F. Carlson Center for Imaging Science located at the Rochester Institute of Technology. Data was collected and preprocessed by Hyperspectives, Inc. This project is funded in part by the United States Air Force Research Laboratory Sensors Directorate.
Project Lead: John Kerekes, Center for Imaging Science, RIT
Webmaster: Stephen Cavilia, Center for Imaging Science, RIT
Development Lead: Kevin King, High School Intern, Center for Imaging Science, RIT
Development Advisor: David Snyder, Center for Imaging Science, RIT
Isaac Gerg, Penn State, manages a forum discussing the Target Detection Blind Test this web application is modeled after. The link is RIT Hyperspectral Target Detection Blind Test Forum.
This project is based on the Target Detection Blind Test developed by David Snyder. The Target Detection Blind Test is described in the following publication:
D. Snyder, J. Kerekes, I. Fairweather, R. Crabtree, J. Shive, and S. Hager, “Development of a Web-based Application to Evaluate Target Finding Algorithms,” Proceedings of the 2008 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), vol. 2, pp. 915-918, Boston, MA, 2008.