You are here
GroundTruth100-for-COIL : 100 labelled images for object recognition tests
To contribute in testing of vision-based object recognition systems, TeV makes available for download a labelled image database, named GroundTruth100-for-COIL, developed for the MEMORI system.
The database consists of 100 synthetic images depicting some objects against a non-uniform background. The objects are rotated and rescaled versions of objects taken from the Columbia Object Image Library COIL-100 (freely available starting from CAVE - Computer Vision Laboratory of Columbia University).
First, the 100 objects of COIL in each one of the 72 poses have been separated from their background using a semi-automatic method. Some of them have been randomly chosen, then rotated with a random angle and scaled with a random factor in the range [0.7, 1.3]. Finally, they have been superimposed on a synthetic, non-uniform, colored background in a non-overlapping fashion.
To evaluate the performance of an object retrieval system, each image is tagged with ground-truth information, i.e. with a labelled map, where each labelled segment corresponds to one object or to background, and with data about the COIL objects present in it: class, pose, rotation angle and scale. Therefore, the GroundTruth100-for-COIL archive comprises three parts: images, labelled maps, data on objects.
To download the GroundTruth100-for-COIL please follow this link where we make also available the masks of the COIL, objects coil-100-masks.
Remark: this data can only be used for research or academic purposes. For a complete test on recognition and pose estimation of objects in images you need COIL-100, too.