HyperLabelMe
A benchmark system for remote sensing image classification
A benchmark system for remote sensing image classification
The Image and Signal Processing (ISP) group at the Universitat de València, http://isp.uv.es, has harmonized a big database of labeled multi- and hyperspectral images for testing classification algorithms. We think that, like in other related fields of science, data sharing and reproducibility are the only ways for fostering true advance in remote sensing data processing. So far we have harmonized 43 image datasets, both multi- and hyperspectral. We want to expand this database as much as possible in order to objectively evaluate algorithms in a common framework.
We provide training pairs (spectra and their labels) and test spectra (test labels are used to assess your predictions). Under any circumstance we are planning to (re)distribute images, only a reduced number of selected pixels. With this platform, researchers are able to train their algorithms locally, and then evaluate their accuracy over an independent, fixed, spectra test set per image. The system returns accuracy and robustness measures of your algorithm in that test set, as well as a ranked list of the best methods.
