SHREC 2016: Track on Partial Shape Queries for 3D Object Retrieval

Overview

Despite numerous recent efforts, 3D object retrieval based on partial shape queries remains a challenging problem, far from being solved. The problem can be defined as: given a partial view of a shape as query, retrieve all partially similar 3D models from a repository. The objective of this track is to evaluate the performance of partial 3D object retrieval methods, for partial shape queries of various qualities and degrees of partiality.

Datasets

The dataset used for evaluation is related to the cultural heritage domain and consists of 3D pottery models originating from the Virtual Hampson Museum collection (http://hampson.cast.uark.edu). The dataset consists of 383 models classified to 6 distinct geometrically defined classes (bottle, bowl, jar, effigy, lithics and others).

The partial shape queries are provided in 3 different forms: (i) artificial queries created by slicing and cap filling of the complete 3D models, (ii) real high quality queries, obtained with the smartSCAN Breuckmann scanner, (iii) real low quality queries, obtained with Microsoft Kinect V2 sensor.

In the case of (ii) and (iii), the queries were created by scanning derivative vessels that have been constructed by a professional potter using the Hampson models as a template. For all 3 forms, queries are provided for multiple degrees of partiality.

Task and procedure

Prospective participants are expected to:

· send an e-mail to ipratika(at)ee.duth.gr until the registration deadline (February 8, 2016), in order to confirm their participation. All group members and their affiliations should be indicated,

· submit a distance matrix for each retrieval setting. Each such setting is associated with one of the 3 different forms of queries, with a specified degree of partiality. Each matrix might be the result of a different algorithm, or a different parameter setting. It is not mandatory for each group to address all 3 forms of queries. Detailed information on the distance matrix file format will be provided soon,

· submit one or many runs.

· provide a one-page description of the testing method.

Each method will be evaluated based on its distance matrices, using standard retrieval scores (nearest neighbor, first tier, second tier, etc).

The track results will be presented into a joint paper, to be published in the proceedings of EG 3DOR 2016.

Timeline

January 13 – Call for participation

January 25 – Sample partial 3D queries and targets will be available online

February 8 – Registration deadline

February 12 – Distribution of the 3D partial query set

February 22 – Submission of the results and one-page description

March 15 – Track paper submission for review

April 1 – Camera-ready track paper submission

May 7-8 – EG 3DOR Workshop, including SHREC’ 2016

Organizers

Ioannis Pratikakis - Democritus University of Thrace, Greece

Michalis Savelonas - Democritus University of Thrace, Greece

Fotis Arnaoutoglou - Democritus University of Thrace, Greece

Anestis Koutsoudis - Athena RC, Greece

Theoharis Theoharis - NTNU, Norway