Plushies: Retrieval of non-rigid (textured) shapes from low quality 3D models
General description
Previous SHREC contests on deformable and textured shape retrieval were mainly based on synthetic deformations of base models with addition of shape and texture perturbations. We propose a real dataset for testing non-rigid and also textured shape retrieval based on low quality acquisitions (based on registration of point clouds acquired with a depth sensor) of a set of plush toys in different poses and with different illumination conditions.Dataset creation
Models have been created by registering (thanks to external landmarks) 4 different point clouds acquired with a Asus Xtion Live Pro depth sensor. Poses are varied making the toys lying on different sides and moving articulated parts. Illumination is varied putting lights in different positions around the models. To allow an easy application of different descriptors, we provide watertight meshes, without texture or with interpolated color values stored as vertex features, saved in OFF format.Data
The first non-textured dataset DATA_NOTEX can be downloaded following this link It includes also the .cla file for testing and text file with indexes of the subsets of pose-varied and illumination-varied models.The textured dataset DATA_TEX1 has been obtained by using a texture parametrization approach and can be downloaded following this link. Note that indexes of corresponding models does not match those of DATA_NOTEX. It includes also the .cla file for testing and text file with indexes of the subsets of pose-varied and illumination-varied models.
The textured dataset DATA_TEX2 has been obtained by using a proximity rule and can be downloaded following this link. It includes also the .cla file for testing and text file with indexes of the subsets of pose-varied and illumination-varied models. Indexes of corresponding models are the same of DATA_TEX1
Results
Preliminary report with the contest results is available from this link.Credits
Data can be freely used for research, provided that any publication derived will cite this site and the paper:Andrea Giachetti, Francesco Farina, Francesco Fornasa, Atsushi Tatsuma, Chika Sanada, Masaki Aono, Silvia Biasotti, Andrea Cerri, Sungbin Choi, SHREC15 Track: Retrieval of non-rigid (textured) shapes using low quality 3D models proc. Eurographics Workshop on 3D Object Retrieval (2015), Zurich, May 2-3, 2015