Overview of benchmark tracks, datasets, and tasks

© 2024 Ben Veldhuijzen, © 2026 Remco C. Veltkamp
Total Tracks: 109
Years Covered: 2006 – 2025
Year Track Title Organizers Reference Database Size Downloads Data Type & Acquisition Category / Task
2006 3D Shape Retrieval Contest Remco C. Veltkamp, Remco Ruijsenaars, Michela Spagnuolo, Roelof van Zwol, Frank ter Haar Veltkamp et al. (2006) 1,814 models
(30 queries)
1,814 models with 3 relevance scores (0-2). Acquired from the Princeton Shape Benchmark. Generic
Classification
2007 Watertight Models Daniela Giorgi, Silvia Biasotti, Laura Paraboschi
daniela.giorgi@ge.imati.cnr.it
silvia.biasotti@ge.imati.cnr.it
Giorgi et al. (2007), in Veltkamp and ter Haar (2007) 400 watertight mesh models Unavailable 400 models divided into 20 classes. Scraped from Drexel, AIM@SHAPE, PSB, CAESAR, McGill, INRIA, etc. Each model is queryied against the rest of the database. Ground truth was established manually. Watertight
Retrieval
2007 CAD Models Yagnanarayanan Kalyanaraman, Karthik Ramani
shrec@purdue.edu
Kalyanaraman & Ramani (2007) in Veltkamp and ter Haar (2007) Unavailable Triangulated meshes of CAD parts in vendor-neutral formats. Sourced from Purdue Engineering Shape Benchmark. The query set was developed for testing search scenarios in: subdivided/decimated models, parmetric variation in the model, slightly modified models and partial shape. CAD
Retrieval
2007 Partial Matching Simone Marini, Laura Paraboschi, Silvia Biasotti
temerina@informatik.uni-freiburg.de
Marini et al. (2007) in Veltkamp and ter Haar (2007) 400 target models
30 partial queries
Unavailable Watertight meshes (.off format). Queries generated by cutting, scaling, and rotating target models. Partial
Retrieval
2007 Protein Retrieval Maja Temerinac, Marco Reisert, Hans Burkhardt
temerina@informatik.uni-freiburg.de
Temerinac et al. (2007) in Veltkamp and ter Haar (2007) 685 proteins
(27 SCOP folds)
Data (15.5 MB) 685 protein domains divided into 27 folds. Data sourced from SCOP database. Protein
Classification
2007 3D Face Models Frank B. ter Haar, Remco C. Veltkamp
Frank ter Haar, Remco C. Veltkamp (2007) in Veltkamp and ter Haar (2007) 1,512 face models Unavailable Emotionless face models generated using a morphable face model (75k vertices, 150k faces). Face
Retrieval
2008 Stability on Watertight Models Silvia Biasotti, Marco Attene
silvia.biasotti@ge.imati.cnr.it
marco.attene@ge.imati.cnr.it
Biasotti & Attene (2008) Set A: 1,229 models
Set B: 1,500 models
Unavailable Perturbed watertight models (Gaussian noise, topology noise, sampling pattern variations). Watertight
Retrieval
2008 Classification of Watertight Models Daniela Giorgi, Simone Marini
daniela@ge.imati.cnr.it
simone@ge.imati.cnr.it
Giorgi and Marini (2008) 570 training models
76 queries
Unavailable 3 class levels (coarse: 12, intermediate: 39, fine: 109). Models sourced from SHREC '07, AIM@SHAPE, Utrecht, etc. Watertight
Classification
2008 CAD Models Track Ramanathan Muthuganapathy, Karthik Ramani
rmuthuga@purdue.edu
ramani@purdue.edu
Muthuganapathy & Ramani (2008) 45 query models Dataset (66MB)
Query (4MB)
CAD parts in vendor-neutral formats organized in 3 super-classes and 45 sub-classes from Purdue Benchmark. CAD
Retrieval
2008 Generic Models Ryutarou Ohbuchi
ohbuchi@yamanashi.ac.jp
Ohbuchi et al. (2008)
[pdf]
1,814 models
(2 sets of 30 queries)
Unavailable Test/Train sets of 907 models each from the Princeton Shape Benchmark. Generic
Retrieval
2008 3D Face Scans Frank B. ter Haar, Mohamed Daoudi, Remco C. Veltkamp
fhaar@cs.uu.nl
mohamed.daoudi@telecom-lille1.eu
ter Haar et al. (2008)
[pdf]
427 range scans
(61 subjects)
Unavailable 7 scans per subject with varying facial expressions from the GavabDB laser range dataset. Face
Retrieval
2009 Structural Shape Retrieval of Watertight Models J. Hartveldt, M. Spagnuolo
jhartvel@cs.uu.nl
michi@ge.imati.cnr.it
Godil et al. (2009) 200 models
(10 classes)
OFF format (183 MB) 200 models in 10 main classes, each having 2 subclasses with 10 models each. Watertight
Retrieval
2009 Querying with Partial Models A. Godil, H. Dutagaci
afzal.godil@nist.gov
Dutagaci et al. (2009) 720 target models in 40 classes, 40 queries Data (188.4 MB) 720 target models from NIST benchmark. Queries contain 20 partial cut models and 20 range scans from NextEngine scanner. Partial
Retrieval
2009 Generic Shape Retrieval A. Godil, H. Dutagaci
afzal.godil@nist.gov
Godil et al. (2009) 720 data objects
80 query objects
Data (131.3 MB) 800 3D objects classified into 40 categories in ASCII .off format. Sourced from NIST Generic Shape Benchmark. Generic
Retrieval
2010 Large Scale Retrieval Remco C. Veltkamp, Geert-Jan Giezeman
G.J.Giezeman@uu.nl
Lian et al. (2010) 10,000 shapes (.ply)
40 queries
Unavailable Dataset contains 493 real models and 9,507 randomly generated procedural shapes without color/texture. Large Scale
Retrieval
2010 Robust Shape Retrieval A. M. Bronstein, M. M. Bronstein, U. Castellani, L. J. Guibas, M. Ovsjanikov
mbron@cs.technion.ac.il
Bronstein et al. (2010) 1,171 test shapes
624 train shapes
Unavailable Transformed triangular meshes (isometries, noise, topology changes). Data from TOSCA, Sumner, and PSB collections. Transformations
Retrieval
2010 Feature Detection and Description A. M. Bronstein et al.
mbron@cs.technion.ac.il
Bronstein et al. (2010) 138 meshes
(3 null meshes)
Unavailable 3 null shapes subjected to 45 transformation variations (scaling, holes, noise, downsampling). TOSCA dataset. Feature Detection
Detection
2010 Correspondence Finding A. M. Bronstein et al.
mbron@cs.technion.ac.il
Bronstein et al. (2010) 138 meshes
(3 null meshes)
Unavailable Evaluates correspondence finding between modified mesh shapes and original null shapes. TOSCA dataset. Transformations
Correspondence
2010 Generic 3D Warehouse T. P. Vanamali, A. Godil, H. Dutagaci
afzal.godil@nist.gov
Vanamali et al. (2010) 3,168 models Data (358.6 MB) 3,168 SketchUp (.skp) models across 43 categories, web-crawled from Google 3D Warehouse. Generic
Retrieval
2010 Non-rigid 3D Shape Retrieval Z. Lian, A. Godil
afzal.godil@nist.gov
Lian & Godil (2010) 200 models Data (259.1 MB) 200 watertight meshes in 10 categories (ants, humans, snakes, etc.) modified from McGill Articulated Benchmark. Non-Rigid
Retrieval
2010 Range Scan Retrieval H. Dutagaci, A. Godil, C. P. Cheung
afzal.godil@nist.gov
Dutagaci et al. (2010) 800 target models
117 range scans
Data (473.3 MB) 800 complete 3D models (.off) target set; queries captured from 40 real objects using a Minolta Laser Scanner. Range Scans
Retrieval
2010 Protein Model Classification L. Mavridis, V. Venkatraman, D. W. Ritchie
Mavridis et al. (2010) 1,000 proteins
50 queries
Unavailable 1,000 proteins from 100 CATH superfamilies with sequence data masked to enforce purely structural evaluation. Protein
Classification
2011 Generic Shape Retrieval Helin Dutagaci, Afzal A. Godil
afzal.godil@nist.gov
Dutagaci et al. (2011) 1,000 models
(50 classes)
Unavailable 1,000 models (.off format) with 20 per class. Sourced from NIST Generic Benchmark and Generic 3D Warehouse. Generic
Retrieval
2011 Robust Feature Detection Benchmark A. M. Bronstein, M. M. Bronstein, A. Kovnatsky, R. Litman, A. Zaharescu
mbron@cs.technion.ac.il
Boyer et al. (2011) 56 meshes Unavailable One human null mesh transformed under 11 transformation classes and 5 strength levels. TOSCA dataset. Feature Detection
Detection
2011 Non-rigid 3D Watertight Meshes Zhouhui Lian, Afzal Godil
shrec@nist.gov
Lian et al. (2011) 600 triangle meshes Data (259.1 MB) 600 watertight models split across 30 categories. 30 original models deformed 19 times using skeleton rigs. Non-Rigid
Retrieval
2011 3D Face Models Retrieval Stefan van Jole, Remco C. Veltkamp, Mohamed Daoudi, Ben Amor Boulbaba
mohamed.daoudi@telecom-lille1.eu
Remco.Veltkamp@cs.uu.nl
Veltkamp et al. (2011) 780 scans
(130 masks)
Unavailable 780 scans collected from 130 physical masks using Roland and Escan 3D scanner systems. Face
Retrieval
2012 3D Mesh Segmentation Guillaume Lavoué et al.
Lavoué et al. (2012) 28 models
112 ground-truths
Unavailable 28 watertight manifold triangle meshes across 5 classes with 4 manual segmentations per mesh from 36 human volunteers. Segmentation
Retrieval
2012 Sketch-Based 3D Shape Retrieval Afzal A. Godil
afzal.godil@nist.gov
Li et al. (2012) 400 3D models
262 sketches
Data (117.9 MB) Target set of 400 watertight models (WMB dataset); 250 hand-drawn sketches and 12 standard line drawings. Sketch
Retrieval
2012 Generic 3D Shape Retrieval Afzal A. Godil
afzal.godil@nist.gov
Li et al. (2012) 1,200 meshes
(60 classes)
Data (133.6 MB) 1,200 triangle meshes (20 per class) compiled from previous SHREC generic datasets, WMB, and PSB. Generic
Retrieval
2012 Stability on Abstract Shapes Silvia Biasotti et al.
Biasotti et al. (2012) 504 meshes 504 watertight meshes synthesized from 18 mathematical primitive shapes perturbed with 9 transformation types. Transformations
Classification
2013 Large-Scale Partial Shape Retrieval I. Sipiran, R. Meruane, B. Bustos, T. Schreck
iasipiranm@gmail.com
Sipiran et al. (2013) 360 target models
7,200 queries
Data (800 MB) 360 target shapes from SHREC '09. 7,200 queries generated by simulating range scans via icosahedral projection planes. Range Scans
Retrieval
2013 Low-Cost Depth-Sensing Cameras Joao Machado, Alfredo Ferreira
alfredo.ferreira@ist.utl.pt
joaoprmachado@gmail.com
Machado et al. (2013) 192 target models
12 query models
Unavailable Low-fidelity watertight meshes scanned using Microsoft Kinect sensors and ReconstructMe software. Range Scans
Retrieval
2013 Retrieval of Textured 3D Models Andrea Cerri, Silvia Biasotti
andrea.cerri@ge.imati.cnr.it
silvia.biasotti@ge.imati.cnr.it
Cerri et al. (2013) 240 watertight meshes
(10 classes)
Unavailable Textured watertight meshes (6 base models per class with 3 textures each), modified with 4 transformation types. Textured
Retrieval
2013 Large Scale Sketch-Based Retrieval Bo Li, Yijuan Lu, Afzal Godil, Tobias Schreck
sketch@nist.gov
li.bo.ntu0@gmail.com
Li et al. (2013) 1,258 3D models
7,200 sketch queries
Data (602.9 MB) 7,200 human sketches (Eitz dataset) and 1,258 3D models from PSB spanning 90 categories. Sketch
Retrieval
2014 Automatic Location of Landmarks used in Manual Anthropometry A. Giachetti, E. Mazzi, F. Piscitelli
andrea.giachetti@univr.it
Godil & Li (2014)
[pdf]
50 train / 50 test Train/Test (98.6 MB) Human body models annotated with manual landmark locations (e.g., Acromiale, Radiale, Stylion). Landmark
Retrieval
2014 Shape Retrieval of Non-Rigid 3D Human Models David Pickup, Xianfang Sun, Paul L Rosin, Ralph R Martin, Zhouhui Lian, Zhiquan Cheng
d.pickup(at)cs.cf.ac.uk
Pickup et al. (2024)
[pdf]
400 (real), 300 (synthetic) - The real dataset is from point-clouds contained within the Civilian American and European Surface Anthropometry Resource (CAESAR), synthetic dataset created in DAZ Studio Non-rigid
Retrieval
2014 Retrieval and classification on Textured 3D Models Andrea Cerri, Silvia Biasotti
andrea.cerri(at)ge.imati.cnr.it and silvia.biasotti(at)ge.imati.cnr.it
Biasotti et al. (2014)
[pdf]
572 watertight mesh models, 16 geometric classes - Synthetic textured mesh models Textured
Retrieval
2014 Extended Large Scale Sketch-Based 3D Shape Retrieval Bo Li, Yijuan Lu, Chunyuan Li, Afzal Godil, Tobias Schreck
sketch(at)nist.gov and b_l58(at)txstate.edu
Li et al. (2014)
[pdf]
8987 3D models in 171 classes, 12680 sketches SHREC2014_SBR.zip (1.03 GB) Generic, articulated, CAD, architecture models, human-drawn sketches Sketch
Retrieval
2014 Large Scale Comprehensive 3D Shape Retrieval Bo Li, Yijuan Lu, Chunyuan Li, Afzal Godil, Tobias Schreck
Generic3D(at)nist.gov and b_l58(at)txstate.edu
Li et al. (2014)
[pdf]
8987 3D models, 171 classes Data (877.5 MB) 8,987 triangle meshes, manual classification Large scale
Retrieval
2015 Canonical Forms for Non-Rigid 3D Shape Retrieval David Pickup, Xianfang Sun, Paul L. Rosin, Ralph R. Martin, Zhiquan Cheng
pickupd@cardiff.ac.uk
Pickup et al. (2015)
[pdf]
100 train, 100 test Data (81.5MB) Canonical forms to factor out a shape’s pose, for rigid retrieval systems to retrieve non-rigid shapes/td> Non-rigid
Retrieval
2015 Non-rigid 3D Shape Retrieval Zhouhui Lian
lianzhouhui@pku.edu.cn
Lian et al. (2015)
[pdf]
1200 watertight triangle meshes, 50 categories Data (241 MB) Models in same category generated by transforming original mesh Non-Rigid
Retrieval
2015 Scalability of Non-Rigid 3D Shape Retrieval I. Sipiran
sipiran@dbvis.inf.uni-konstanz.de
Sipiran et al. (2015)
[pdf]
96487 models, 9 classes - Similated range images Non-rigid
Retrieval
2015 3D Object Retrieval with Multimodal Views Yue Gao
kevin.gaoy@gmail.com
Gao et al. (2025)
[pdf]
505 objects, 311 queries, 73 images+depth / object - Recorded by Kinect sensors from 3 directions Textured/Multiview
Retrieval
2015 Retrieval of non-rigid (textured) shapes using low quality 3D models Andrea Giachetti, Francesco Fornasa, Francesco Farina
andrea.giachetti@univr.it
Giachetti et al. (2015)
[pdf]
240 textured, 120 non-textured of 12 toys Data (33.4MB) Point cloud meshed to create the watertight models, scanned with Asus Xtion Live Pro for depth and texture Non-rigid/Textured
Retrieval
2015 Retrieval of Objects Captured with Kinect One Camera Pedro B. Pascoal, Pedro Proença, Miguel Sales Dias, Alfredo Ferreira
t-pedrop@microsoft.com
Pascoal et al. (2015)
[pdf]
175 objects in 18 classes - Household objects with Kinect 1 Range scans
Retrieval
2015 Range Scans based 3D Shape Retrieval A. Godil
afzal.godil@nist.gov
Godil et al. (2015)
[pdf]
1200 target, 180 query Data (655.4 MB) 60 object, Minolta scanner Range scans
Retrieval
2016 Retrieval of Human Subjects from Depth Sensor Data Andrea Giachetti
andrea.giachetti@univr.it
Giachetti et al. (2016)
[pdf]
720 testing, 180 training Train (93 MB), test (319 MB) 18 scans for each subject, 40 test subjects, 10 training subjects Range scans (human)
Retrieval
2016 3D Sketch-Based 3D Shape Retrieval Bo Li, Yijuan Lu
li.bo.ntu0@gmail.com
Li et al. (2016)
[pdf]
300 3D sketches in 30 classes, 1258 models in 90 classes Data (71 MB) 3D sketches where created with a Kinect Sketch
Classification
2016 Matching of Deformable Shapes with Topological Noise Zorah Lähner, Emanuele Rodolà, Michael Bronstein, Daniel Cremers
laehner@in.tum.de
Lähner et al. (2016)
[pdf]
25 shapes (15 for training 10 for testing) - The (human) shapes are deformable and have topological noise, created with DAZ 3D studio Non-rigid
Retrieval
2016 Partial Matching of Deformable Shapes Luca Cosmo, Emanuele Rodolà, Michael Bronstein, Andrea Torsello
shrec2016@dais.unive.it
Cosmo et al. (2016)
[pdf]
76 base shapes in 8 classes, 596 total shapes - Remeshed to 10K vertices, made partial either by adding holes randomly (276 shapes) or cutting the shape in a random direction (320 shapes), base shapes are from the TOSCA dataset Partial
Retrieval
2016 Shape Retrieval of Low-Cost RGB-D Captures Pedro Pascoal, Pedro Proença, Miguel Sales Dias, Alfredo Ferreira
pmbp@tecnico.ulisboa.pt
Pascoal et al. (2016)
[pdf]
200 models - 90 frame pairs of RGB and Depth images, models from the 3D SketchUp Warehouse Range scans
Classification
2016 Partial Shape Queries for 3D Object Retrieval Ioannis Pratikakis, Michalis Savelonas, Fotis Arnaoutoglou, Anestis Koutsoudis, Theoharis Theoharis
ipratika@ee.duth.gr
Pratikakis et al. (2016)
[pdf]
383 models in 6 classes - 3 qualities: artificial (slicing and capfilling), real queries high quality (smartSCAN), real queries low quallity (Kinect) Partial
Retrieval
2016 Large-Scale 3D Shape Retrieval from ShapeNet Core55 Manolis Savva, Fisher Yu, Hao Su
shrec2016shapenet@gmail.com
Savva et al. (2016)
[pdf]
51190 3D models, 70/10/20% training / validation / test - Categorized into 55 WordNet categories and 204 sub-categories, models where deduplicated Scalability
Retrieval
2016 3D Object Retrieval with Multimodal Views Yue Gao, Anan Liu, Weizhi Nie, Qionghai Dai
weizhinie@tju.edu.cn
Gao et al. (2016)
[pdf]
605 objects, 200 queries, 405 test/target in 60 categories Download (size) 100 3D printed and 505 real 3D objects scanned with Kinect from 3 angles Range scans/Multiview
Retrieval
2017 RGB-D to CAD Retrieval with ObjectNN Dataset Binh-Son Hua, Quang-Hieu Pham, Minh-Khoi Tran, Quang-Trung Truong Hua et al. (2017)
[pdf]
4975 models, 50 / 25 / 25 train / validation / test Data (3,8 GB), extras on Github 1667 SceneNN objects, 3308 ShapeNet models into 20 categories CAD
Retrieval
2017 3D Hand Gesture Recognition Using a Depth and Skeletal Dataset Quentin De Smedt, Hazem Wannous, Jean-Phillipe Vandeborre
david.filliat@ensta-paristech.fr
De Smedt et al. (2017)
[pdf]
2800 sequences, 14 gestures Data (5,8 GB) sequences of 14 hand gestures performed multiple times with hand or finger Gesture
Classification
2017 Large-Scale 3D Shape Retrieval from ShapeNet Core55 Manolis Savva, Hao Su, Fisher Yu, Thomas Funkhouser
shrecshapenet@gmail.com
Savva et al. (2017)
[pdf]
51162 3D models in 55 WordNet categories, 204 sub-categories Data (50 GB) models where deduplicated Scalability
Retrieval
2017 Protein Shape Retrieval Haiguang Liu Song et al. (2017)
[pdf]
5854 target, 10 queries - low quality models removed from original set Protein
Retrieval
2017 Point-Cloud Shape Retrieval of Non-Rigid Toys Frederico A. Limberger, Richard C. Wilson
pronto-group@york.ac.uk
Limberger et al. (2017)
[pdf]
100 3D scanned models, 10 classes Data (3,6 MB) 10 poses per model, scanned by Head & Face Color 3D Scanner Range scan
Classification
2017 Deformable Shape Retrieval with Missing Parts Emanuele Rodolà, Or Litany, Michael Bronstein Rodolà et al. (2017)
[pdf]
1216 train (holes), 1078 test (holes), 1082 train (range), 882 test (range) - Non-rigid deformations, different amounts and types of partiality, topological changes induced by mesh gluing in areas of contact Partial
Retrieval
2017 Retrieval of surfaces with similar relief patterns Silvia Biasotti, Andrea Giachetti
andrea.giachetti@univr.it
Biasotti et al. (2017)
[pdf]
720 models in 15 classes - 180 models + 3 tesselation variations of each model Relief
Classification
2018 2D Sketch-Based 3D Scene Retrieval Juefei Yuan, Bo Li, Yijuan Lu
bo.li@usm.edu
Yuan et al. (2018)
[pdf]
250 Scene sketches, 1000 3D scene models in 10 classes Data (4,5 GB) 180 train sketches, 70 test sketches, 700 train 300 test 3d models, 3D models from Google 3D warehouse Sketch
Retrieval
2018 2D Image-Based 3D Scene Retrieval Hameed Abdul-Rashid, Juefei Yuan, Bo Li, Yijuan Lu
bo.li@usm.edu
Abdul-Rashid et al. (2018)
[pdf]
10.000 2D images, 1000 3D models in 10 classes Data (5,6 GB) 180 train images, 70 test images, 700 train, 300 test 3d models, 3D models from Google 3D warehouse Image
Retrieval
2018 RGB-D Object-to-CAD Retrieval Quang-Hieu Pham, Binh-Son Hua
quanghieu_pham@mymail.sutd.edu.sg
Pham et al. (2018)
[pdf]
2101 query 3308 target - Real-world SceneNN and ScanNet target dataset, subset of ShapeNetSem CAD
Retrieval
2018 Protein Shape Retrieval Matthieu Montes, Florent Langenfeld
matthieu.montes@cnam.fr
Langenfeld et al. (2018)
[pdf]
2267 protein structures in 107 classes Data (3,1 GB) Use each member of class as query for the rest of the class Protein
Retrieval
2018 Retrieval of grey patterns depicted on 3D models E. Moscoso Thompson, S. Biasotti
elia.moscoso@ge.imati.cnr.it, silvia.biasotti@ge.imati.cnr.it
Moscoso Thompson et al. (2018)
[pdf]
300 surfaces 20 base models sample (10 MB) patterns applied to the meshes of basic shapes (cube, cylinder, cup, vase) Textured
Classification
2018 Recognition of geometric patterns over 3D models E. Moscoso Thompson, S. Biasotti, G. Sorrentino, M. Polig, S. Hermon
elia.moscoso@ge.imati.cnr.it, silvia.biasotti@ge.imati.cnr.it
Biasotti et al. (2018)
[pdf]
8 query meshes, 6 patterns 30 models - 25 models have at least 1 pattern, archeological artefacts from EU H2020 project GRAVITATE Relief
Retrieval
2019 Extended 2D Scene Sketch-Based 3D Scene Retrieval Juefei Yuan, Hameed Abdul-Rashid, Bo Li, Yijuan Lu, Tobias Schreck
juefei.yuan@usm.edu
Yuan et al. (2019)
[pdf]
750 2D scene sketches, 3000 3D scene models in 30 classes Data (16,3 GB) per class 18 train and 7 test sketches, 70 train and 30 test models Sketch
Retrieval
2019 Extended 2D Scene Image-Based 3D Scene Retrieval Hameed Abdul-Rashid, Juefei Yuan, Bo Li, Yijuan Lu, Tobias Schreck
hameed.abdulrashid@usm.edu
Abdul-Rashid et al. (2019)
[pdf]
30000 2D, 3000 3D in 30 classes Data (16,3 GB) Images from ImageNet, models from 3D Warehouse Image
Retrieval
2019 Feature Curve Extraction on Triangle Meshes E. Moscoso Thompson
elia.moscoso@ge.imati.cnr.it
Moscoso Thompson et al. (2019)
[pdf]
15 surfaces Samples (4,1 MB) Models scanned or made in silico, derived from Turbosquid repository of 3D models or the Visionair shape workbench Detection
Retrieval
2019 Protein Shape Retrieval Florent Langenfeld, Matthieu Montes
matthieu.montes@cnam.fr
Langenfeld et al. (2019)
[pdf]
5298 modelsin 17 classes Species proteins Off-files (11.6 GB) From the SCOPe database, then randomly selected to decrease to 5298 models Protein
Retrieval
2019 Classification in Cryo-Electron Tomograms lja Gubins, Gijs van der Schot, Remco C. Veltkamp, Friedrich G. Forster
i.gubins@uu.nl
Gubins et al. (2019)
[pdf]
10 tomograms, 2540 proteins, 12 classes Data (6,5 GB) 10 reconstructed tomograms obtained from simulated cell-like volume, each filled with on average 2500 non-overlapping proteins. Protein
Classification
2019 Online Gesture Recognition Fabio Marco Caputo, Andrea Giachetti
fabiomarco.caputo@univr.it
Caputo et al. (2019)
[pdf]
training set of 60 recordings from 4 subjects, test set of 135 recordings, 5 gestures Train (8.9 MB) Recordings done in VR Oculus Rift in interactive setting Gesture
Classification
2019 Monocular Image-based 3D Model Retrieval An-An Liu, Wei-Zhi Nie, Wen-Hui Li, Dan Song, Yu-Qian Li, He-Yu Zhou, Shu Xiang, Wei-Jie Wang
liwenhui@tju.edu.cn
Li et al. (2019)
[pdf]
21000 2D images, 7690 3D objects, 12 views/object, 21 classes - 2D images collected from ImageNet, 3D objects collected from NTU, PSB, ModelNet40, ShapeNet Image-based
Retrieval
2019 Shape Correspondence with Isometric and Non-Isometric Deformations Roberto Dyke, Caleb Stride, Yu-Kun Lai, and Paul L. Rosin
DykeRM@cardiff.ac.uk
Dyke et al. (2019)
[pdf]
76 shape pairs Data (53.1 MB) Artec3D Space Spider scanner, articulating, bending, stretching, topological deformations Transformations
Correspondence
2019 Correspondence in Humans with Different Connectivity Simone Melzi, Riccardo Marin, Emanuele Rodolà , Umberto Castellani
simone.melzi@univr.it
Melzi et al. (2019)
[pdf]
430 shape pairs Data (48.9 MB) Models from SMPL, FAUST, SCAPE, TOSCA, SPRING, MoSh Mocap, Princeton, CEASAR, SHREC14 Non-Rigid 3D Human, and others Humans
Correspondence
2020 Non-rigid Shape Correspondence of Physically-Based Deformations Roberto Dyke, Feng Zhou, Yu-Kun Lai, and Paul L. Rosin
DykeRM@cardiff.ac.uk
Dyke et al. (2020)
[pdf]
11 partial scans and 1 full scan of rabit Data (13 MB) Artec3D Space Spider scans of soft stretchy toy rabbit, with stretch, indent, twist, inflate deformations Transformation
Correspondence
2020 Shape correspondence with non-isometric deformations Roberto Dyke, Yu-Kun Lai, and Paul L. Rosin
DykeRM@cardiff.ac.uk
Dyke et al. (2020) 14 animal models Data (10 MB) Meshes (simplified to) 100.000 traingles Transformation
Correspondence
2020 3D point cloud semantic segmentation for street scenes Tao Ku, Remco C. Veltkamp
t.ku@uu.nl
Ku et al. (2020) 60 training, 20 test 3D point cloud for street scenes Drive (2,6 GB) Scenes from Cyclomedia Technology with panoramic camera and velodyne HDL-32 Lidar sensor, manually labelled, two million points per point cloud Range scans
Classification
2020 Multi-domain protein shape retrieval challenge Matthieu Montès, Florent Langenfeld
matthieu.montes@cnam.fr
Langenfeld et al. (2020) 588 protein chains from 26 species Website (1.2 GB) SCOPe database en PDB (Protein Data Bank), classes at least 10 members, duplicates removed Protein
Classification
2020 Retrieval of digital surfaces with similar geometric reliefs Elia Moscoso Thompson, Silvia Biasotti, Andrea Giachetti
elia.moscoso@ge.imati.cnr.it
Moscoso Thompson et al. (2020) 220 surfaces Data (227 MB) Relief patterns applied to randomly rotated models Relief
Classification
2020 Classification in cryo-electron tomograms Ilja Gubins, Marten Chaillet, Gijs van der Schot, Remco Veltkamp, Friedrich Förster
i.gubins@uu.nl
Gubins et al. (2020) 9 tomograms, between 2400 and 2800 proteins, 12 classes Data (7,2 GB)) Sphericity, radius, electron density maps generated Protein
Classification
2020 River gravel characterization Andrea Giachetti, Silvia Biasotti, Luigi Fraccarollo, Filippo Andrea Fanni
andrea.giachetti@univr.it
Giachetti et al. (2020)
[pdf]
256 surfaces, 8 classes of grain Data (420 MB) Photogrammetry to create 3D models of the patches Textured
Classification
2020 6D object pose estimation Honglin Yuan, Remco C. Veltkamp
h.yuan@uu.nl
Yuan et al. (2020)
[pdf]
500 PNG 1280x720 image pairs, 400 trainig, 100 test Data Training set of synthesized and testing set of captured synthesized color-and-depth image pairs, objects have size, shape, texture, and reflective characteristics, real-world data captured by Intel RealSense depth camera D415 and data generated from simulation for the 6D object pose Pose
Estimation
2020 Extended Monocular Image-based 3D Object Retrieval Wenhui Li, Dan Song, Anan Liu, Weizhi Nie, Ting Zhang, Xiaoqian Zhao, Mingsheng Ma, Yuqian Li, Heyu Zhou
dan.song@tju.edu.cn, anan0422@gmail.com
Li et al. (2020)
[pdf]
40.000 2D images, 12.732 3D models, 40 categories<, train/test 50/50%/td> - Every category has between 52-500 3D models and 1000 2D images Image-based
Retrieval
2021 Retrieval of cultural heritage objects Ivan Sipiran, Patrick Lazo, Cristian Lopezc, Milagritos Jimenez
isipiran@dcc.uchile.cl
Sipiran et al. (2021) 938 3D models in 8 categories, train/test 70/30% Data (1.8 GB) Remeshed to 40.000 triangle faces, ground truth (shape, and culture) from curators, archaeological objects from Josefina Ramos de Cox Museum in Peru with varied geometry and artistic styles, scanned with structured-light desktop scanner Textured
Retrieval
2021 Retrieval and classification of protein surfaces equipped with physical and chemical properties Andrea Raffoa, Ulderico Fugacci, Silvia Biasotti, Walter Rocchia
andrea.raffo@ge.imati.cnr.it, ulderico.fugacci@ge.imati.cnr.it, silvia.biasotti@ge.imati.cnr.it, walter.rocchia@iit.it
Raffoa et al. (2021) 5000 protein surfaces, 209 PDB entries, train/test 70/30% - Each PDB entry in different shapes experimentally determined via NMR measurements, based on the 2019 SHREC track Protein
Retrieval
2021 Surface-based protein domains retrieval Florent Langenfeld, Matthieu Montes
florent.langenfeld@lecnam.net
Langenfeld et al. (2021)
[pdf]
554 molecular surfaces, 2 sets of 10 queries (with and without electro) Website (3.2 GB) 2 x 10 Pfam domains, surface mesh computed with EDTSurf, task is to compute 2x10 dissimilarity matrices Protein
Retrieval
2021 Classification in Cryo-Electron Tomograms Ilja Gubins, Marten Chaillet, Gijs van der Schot, Remco C. Veltkamp, Friedrich G. Förster
i.gubins@uu.nl
Gubins et al. (2021)
[pdf]
10 tomograms, 512x512x512 1nm/voxel in 12 protein classes Data (7.4 GB) Generated by physics-based simulation, 1000-1300 proteins, 7-14 gold fiducials, 2-7 vesicles at random lacations and orientations without overlap per tomogram Protein
Classification
2021 Skeleton-based Hand Gesture Recognition in the Wild Fabio Marco Caputo, Andrea Giachetti
fabiomarco.caputo@univr.it
Caputo et al. (2021) 180 gesture sequences, 18 gesture classes, 40 occurences/class Training (99MB), Test (59 MB) Captured using a Leap Motion Gesture
Classification
2021 3D Point Cloud Change Detection for Street Scenes Tao Ku, Sam Galanakis, Bas Boom, Remco C. Veltkamp
T.Ku@uu.nl
Ku et al. (2021) 78 large-scale street scene 3D point clouds, 866 object pairs, train/test 82/18% Drive (1.6 GB) Dataset provided by CycloMedia Technology, colored point clouds gathered in 2016 and 2020 in Schiedam with LiDAR sensors on vehicles, manual labeling Point cloud
Detection
2021 Quantifying Shape Complexity Mazlum Ferhat Arslan, Alexandros Haridis, Paul L. Rosin, Sibel Tari
ferhata@metu.edu.tr
Ferhat Arslan et al. (2021) set 1: 1800 models of cubes and spheres, set 2: 50 shapesof cuboids, set 3: 380 shapes Drive (5.2 GB) Set 1 and 2 generated, set 3 from Princeton Mesh Segmentation Benchmark Complexity
Classification
2022 Online detection of heterogeneous gestures Marco Emporio, Anton Pirtac, Ariel Caputo, Marco Cristani, Andrea Giachetti
andrea.giachetti@univr.it
Emporio et al. (2022) 16 classes gestures, 288 sequences, 3-5 gestures per sequence Training (29 MB)Test (29 MB) Captured using the Hololens 2 finger tracking headset in a realistic use-case of mixed reality interaction Gesture
Classification
2022 Fitting and recognition of simple geometric primitives on point clouds Chiara Romanengo, Andrea Raffo
chiara.romanengo@ge.imati.cnr.it
Romanengo et al. (2022) 46.925 point clouds in 5 classes, 46000 train, 925 test Webpage (7.6 GB) Point clouds generated from surfaces, then perturbed in 10 ways Point cloud
Fitting
2022 Open-Set 3D Object Retrieval using Multi-Modal Representation Yue Gao, Yifan Feng, Xibin Zhao, Yandong Guo
evanfeng97@gmail.com
Feng et al. (2022) 12309 objects in 40 classes, 2822 from 8 categories training set, 960 query objects and 8527 target objects from 32 other categories OS-MN40 (46 GB), OS-MN40-Miss (28 GB) Multimodal and multi-resolution, based on ModelNet40, manually split the categories in the train and retrieval set to simulate an open-set, also split into with/without modality. Multimodal
Retrieval
2022 Pothole and crack detection on road pavement using RGB-D images A. Ranieri, E. Moscoso Thompson, S. Biasotti
elia.moscoso@ge.imati.cnr.it
Moscoso Thompson et al. (2022) 4340 image pairs 797 RGB-D videos, 3340 training, 496 validation, 504 test images Mendeley Data (6.39 GB) Image set based on data sets Crack500, GAPs384, EdmCrack600, Pothole-600, Cracks and Potholes in Road Images Dataset, RGB-D video clips made with Luxonis OAK-D camera Relief
Detection
2022 Sketch-Based 3D Shape Retrieval in the Wild Jie Qin, Shuaihang Yuan, Jiaxin Chen, Boulbaba Ben Amor, Yi Fang
qinjiebuaa@gmail.com
Qin et al. (2022) 46000 CAD-based point cloud models, 1700 scanned point cloud models, 145000 sketches Drive Based on QuickDraw sketches, ModelNet40 and ShapeNet Core55 meshes, ScanObjectNN scans Sketch
Retrieval
2022 Protein-ligand binding site recognition Luca Gagliardi, Walter Rocchia, Andrea Raffo, Ulderico Fugacci, Silvia Biasotti
andrea.raffo@ge.imati.cnr.it
Gagliardi et al. (2022) 1091 proteins, 1721 ligand binding sites Github (1.8 GB) PQR file using AMBER force field is created using pdb2pqr software, meshes created with NanoShaper Protein
Recognition
2023 Detection of symmetries on 3D point clouds representing simple shapes Ivan Sipiran, Chiara Romanengo, Silvia Biasotti, Bianca Falcidieno
isipiran@dcc.uchile.cl
Sipiran et al. (2023)
[pdf]
69.000 Pointclouds of simple shapes (60k train, 9k test), 5 perturbations, 7-curve forms Drive (7.9 GB) Curve forms generate cylinders and cones, transformed to point clouds, perturbed with noise Symmetry
Detection
2023 Sketch-based 3D Animal Fine-Grained Retrieval Trung-Nghia Le, Minh-Triet Tran, Minh-Quan Le, Xuan-Nhat Hoang, Thang-Long Nguyen-Ho, Trong-Thuan Nguyen, Trong-Le Do, Vinh-Tiep Nguyen, Tam V. Nguyen, Akihiro Sugimoto
ltnghia@fit.hcmus.edu.vn, tmtriet@fit.hcmus.edu.vn
Le et al. (2023) 711 animal models, 140 sketch queries Drive (229 MB) Collected models are made watertight and reduced, sketches are made manually Sketch
Retrieval
2023 Text-based 3D Animal Fine-Grained Retrieval Trung-Nghia Le, Minh-Triet Tran, Minh-Quan Le, Xuan-Nhat Hoang, Thang-Long Nguyen-Ho, Trong-Thuan Nguyen, Trong-Le Do, Vinh-Tiep Nguyen, Tam V. Nguyen, Akihiro Sugimoto
ltnghia@fit.hcmus.edu.vn, tmtriet@fit.hcmus.edu.vn
Le et al. (2023) 711 animal models, 150 sentences as queries Drive (227 MB) Collected models are made watertight and reduced, text descriptions are made manually Text
Retrieval
2023 Point Cloud Change Detection for City Scenes Honglin Yuan, Yang Gao, Tao Ku, Remco C. Veltkamp
hlyuan@nuist.edu.cn
Gao et al. (2023) 1711 object change pairs, 78 city-scenes, 5 change classes, train/test 80/20% Data (8,94 GB) LiDAR data is from SHREC 2021 track, simulated data created with Unreal Engine 4, labels manually Range scan
Detection
2024 Non-rigid Complementary Shapes Retrieval in Protein-protein Interactions Florent Langenfeld, Matthieu Montes
florent.langenfeld@cnam.fr, matthieu.montes@cnam.fr
Yacoub et al. (2024)
[pdf]
387 query and 520 target surfaces, 52 protein-protein interactions Queries (1 GB),
Targets (1.2 G)
Based on docking benchmark version 5 Protein
Retrieval
2024 Recognition of hand motions molding clay Ben Veldhuijzen, Remco C. Veltkamp
b.veldhuijzen@students.uu.nl
Veldhuijzen et al. (2024) 62 motion sequences (between 29 and 3721 frames) in 7 classes, training/test 70/30% Data Capturing an experienced potter with a Vicon System containing 14 Vantage cameras Gesture
Classification
2025 Partial Retrieval Benchmark Bart Iver van Blokland, Ivan Sipiran, Benjamin Bustos, Silvia Biasotti, Giorgio Palmieri
bart.van.blokland@ntnu.no
van Blokland et al. (2025) dynamic Github ShapeBench generates point clouds dynamically by sampling mesh data from Objaverse-1.0, which contains over 800,000 3D objects Partial
Retrieval
2025 Protein Shape Classification Taher Yacoub, Camille Depenveiller, Matthieu Montès
taher.yacoub@lecnam.net
Yacoub et al. (2025) 11,565 surfaces divided into 97 imbalanced classes, train/test 80/20% Training (4.1 GB), Test (1 GB) Surfaces generated with NanoShaper, the potentials are calculated using APBS Protein
Classification
2025 Retrieval and Segmentation of Multiple Relief Patterns Gabriele Paolini, Claudio Tortorici, Stefano Berretti
gabriele.paolini1@unifi.it
Paolini et al. (2025)
[pdf]
700 training meshes, 300 test meshes, 54 query meshes GitHub 15 base models Relief
Retrieval
2025 3D Object Retrieval & Completion with Gaussian Splatting Organizers Thien-Phuc Tran, Minh-Quang Nguyen, Minh-Triet Tran, Tam V. Nguyen, Minh Do, Trong-Thuan Nguyen, Viet-Tham Huynh
ttphuc21@apcs.fitus.edu.vn
Tran et al. (2025) 542 synthetic 3D objects in 107 semantic classes, 7705 pre-segmented Guassian Splatting clusters Upon request Meshes converted to Gaussian Splatting representations Part-based
Retrieval
2025 Intelligent 3D Room Design Trong-Thuan Nguyen, Viet-Tham Huynh, Minh-Triet Tran, Tam V. Nguyen
ntthuan@selab.hcmus.edu.vn
Nguyen et al. (2025) Over 1,600 apartment scenes, nearly 5,200 rooms, and more than 44,000 targeted queries GitHub 3D models are converted to multiple views and point clouds Layout
Classification
20xx Track Organizers
Contact
Ref DB size Download (size) Data Type & Acquisition Category
Task