Datasets Images for testing computer vision algorithms

Models

Artificially generated models of objects of two types: cuboids and cylinders, stored as dense colour point clouds, supplemented with sparse clouds of features (currently SIFTs).

Those models were generated for the purpose of testing object recognition methods (with high quality models) and may also be used as ground truth for algorithms of automatic registration of object models from different viewpoints.

Dataset contains also high resolution images from which the models were generated: six images for each model of cuboidal objects and three images for cylindrical objects, along with the YML files containing description of their geometry and mappings of images to model faces.


File description

faces
model textures (6 faces for cuboids, 3 for cylinders)
model
pointclouds with resolution 1pt/1mm; *_xyzrgb - dense RGB cloud; *_xyzsift - sparse cloud of SIFT features
model2
pointclouds with resolution 2pt/1mm; *_xyzrgb - dense RGB cloud; *_xyzsift - sparse cloud of SIFT features