3DOR 15

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/12281

Browse

Now showing 1 - 20 of 23

3D GrabCut: Interactive Foreground Extraction for Reconstructed 3D Scenes
(The Eurographics Association, 2015) Meyer, Gregory P.; Do, Minh N.; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
In the near future, mobile devices will be able to measure the 3D geometry of an environment using integrated depth sensing technology. This technology will enable anyone to reconstruct a 3D model of their surroundings. Similar to natural 2D images, a 3D model of a natural scene will occasionally contain a desired foreground object and an unwanted background region. Inspired by GrabCut for still images, we propose a system to perform interactive foreground/background segmentation on a reconstructed 3D scene using an intuitive user interface. Our system is designed to enable anyone, regardless of skill, to extract a 3D object from a 3D scene with a minimal amount of effort. The only input required by the user is a rectangular box around the desired object. We performed several experiments to demonstrate that our system produces high-quality segmentation on a wide variety of 3D scenes.
3D Object Retrieval with Multimodal Views
(The Eurographics Association, 2015) Gao, Yue; Liu, Anan; Nie, Weizhi; Su, Yuting; Dai, Qionghai; Chen, Fuhai; Chen, Yingying; Cheng, Yanhua; Dong, Shuilong; Duan, Xingyue; Fu, Jianlong; Gao, Zan; Guo, Haiyun; Guo, Xin; Huang, Kaiqi; Ji, Rongrong; Jiang, Yingfeng; Li, Haisheng; Lu, Hanqing; Song, Jianming; Sun, Jing; Tan, Tieniu; Wang, Jinqiao; Yin, Huanpu; Zhang, Chaoli; Zhang, Guotai; Zhang, Yan; Zhang, Yan; Zhao, Chaoyang; Zhao, Xin; Zhu, Guibo; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
This paper reports the results of the SHREC'15 track: 3D Object Retrieval with Multimodal Views, which goal is to evaluate the performance of retrieval algorithms when multimodal views are employed for 3D object representation. In this task, a collection of 505 objects is generated and both the color images and the depth images are provided for each object. 311 objects are selected as the queries and average retrieval performance is measured. The track attracted six participants and the submission of 26 runs, to two tasks. The evaluation results show a promising scenario about multimodal view-based 3D retrieval methods, and reveal interesting insights in dealing with multimodal data.
3D Object Retrieval with Parametric Templates
(The Eurographics Association, 2015) Getto, Roman; Fellner, Dieter W.; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
We propose a 3D object retrieval system which uses parametric templates as prior knowledge for the retrieval. A parametric template represents an object-domain and a semantic concept like 'chair' or 'plane' or a more specific concept like 'dining-char' or 'biplane'. The template can be specified at a general or specific level and can even equal actual retrieved objects. The parametric template is composed of several input parameters and an operation chain which constructs an object. Different parameter combinations lead to different object instances. We combine and evaluate a paramteric template with different descriptors. Our results show that the usage of parametric templates can raise the retrieval performance significantly.
Accelerating the Computation of Canonical Forms for 3D Nonrigid Objects using Multidimensional Scaling
(The Eurographics Association, 2015) Shamai, Gil; Zibulevsky, Michael; Kimmel, Ron; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
The analysis of 3D nonrigid objects usually involves the need to deal with a large number of degrees of freedom. When trying to match two such objects, one approach is to map the surfaces into a domain in which the matching process is simple to execute. Limiting the discussion to almost isometric mappings, which describe most natural deformations in nature, one could resort to Canonical forms. Such forms translate the surface's intrinsic geometry into an extrinsic one in a Euclidean space, thus eliminating the effect of deformations at the expense of (hopefully) minor embedding errors. Multidimensional Scaling (MDS) is a dimensionality reduction technique that can be used to compute canonical forms of 3D-objects, by first evaluating the pairwise geodesic distances between surface points, and then embedding the distances in a lower dimensional Euclidean space. The native computational and space complexities involved in describing such inter-geodesic distances is quadratic in the number of surface points, a property that could be prohibiting in various scenarios. We present an acceleration framework for multidimensional scaling, by accurately approximating the pairwise distance maps. We show how the proposed Nyström Multidimensional Scaling (NMDS) framework can be used to compute canonical forms in quasi-linear time and linear space complexities in the number of data points. It allows us to efficiently deal with high resolution structures without giving up the embedding accuracy.
Automatic 3D Object Fracturing for Evaluation of Partial Retrieval and Object Restoration Tasks - Benchmark and Application to 3D Cultural Heritage Data
(The Eurographics Association, 2015) Gregor, Robert; Bauer, Danny; Sipiran, Ivan; Perakis, Panagiotis; Schreck, Tobias; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Recently, 3D digitization and printing hardware have seen rapidly increasing adoption. High-quality digitization of real-world objects is becoming more and more efficient. In this context, growing amounts of data from the cultural heritage (CH) domain such as columns, tombstones or arches are being digitized and archived in 3D repositories. In many cases, these objects are not complete, but fragmented into several pieces and eroded over time. As manual restoration of fragmented objects is a tedious and error-prone process, recent work has addressed automatic reassembly and completion of fragmented 3D data sets. While a growing number of related techniques are being proposed by researchers, their evaluation currently is limited to smaller numbers of high-quality test fragment sets. We address this gap by contributing a methodology to automatically generate 3D fragment data based on synthetic fracturing of 3D input objects. Our methodology allows generating large-scale fragment test data sets from existing CH object models, complementing manual benchmark generation based on scanning of fragmented real objects. Besides being scalable, our approach also has the advantage to come with ground truth information (i.e. the input objects), which is often not available when scans of real fragments are used. We apply our approach to the Hampson collection of digitized pottery objects, creating and making available a first, larger restoration test data set that comes with ground truth. Furthermore, we illustrate the usefulness of our test data for evaluation of a recent 3D restoration method based on symmetry analysis and also outline how the applicability of 3D retrieval techniques could be evaluated with respect to 3D restoration tasks. Finally, we discuss first results of an ongoing extension of our methodology to include object erosion processes by means of a physiochemical model simulating weathering effects.
Bag of Compact HKS-based Feature Descriptors
(The Eurographics Association, 2015) ElNaghy, Hanan; Hamad, Safwat; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
3D object retrieval has become an integral part in many today's applications attracting extensive research efforts. This paper introduces an enhanced 3D object retrieval technique using a compact and highly discriminative feature point descriptor. The key idea is based on integrating Bag of features (BoF) paradigm with Heat Kernel Signature (HKS) for feature description and detection. Initially, HKS computation phase defines HKS point signatures for each 3D model. Then, an innovative feature point detection algorithm provides a succinct set of feature points to be associated with a compact HKS-based descriptor vectors computed at local time scales. Finally, we take advantage of the BoF paradigm to encode a given 3D model with an informative feature frequency vector. The proposed approach has been evaluated on SHREC 2015 dataset of non-rigid models. The experimental results demonstrate the effective retrieval performance, invariance to different kinds of deformation and possible noise.
Canonical Forms for Non-Rigid 3D Shape Retrieval
(The Eurographics Association, 2015) Pickup, David; Sun, Xianfang; Rosin, Paul L.; Martin, Ralph R.; Cheng, Zhiquan; Nie, Sipin; Jin, Longcun; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
We present a new benchmark for testing algorithms that create canonical forms for use in non-rigid 3D shape retrieval. We have combined two existing datasets to create a varied collection of models for testing. Canonical forms attempt to factor out a shape's pose, giving a pose-neutral shape. This opens up the possibility of using methods originally designed for rigid retrieval for the task of non-rigid shape retrieval. We demonstrate the benchmark by using it to compare the performance of nine canonical form methods, using three different retrieval algorithms.
Computing Local Binary Patterns on Mesh Manifolds for 3D Texture Retrieval
(The Eurographics Association, 2015) Werghi, Naoufel; Tortorici, Claudio; Berretti, Stefano; Bimbo, Alberto Del; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
In this paper, we present and experiment a novel approach for retrieving 3D geometric texture patterns on 2D mesh-manifolds (i.e., surfaces in the 3D space) using local binary patterns (LBP) constructed on the mesh. The method is based on the recently proposed mesh-LBP framework [WBD15]. Compared to its depth-image counterpart, the mesh-LBP is distinguished by the following features: a) inherits the intrinsic advantages of mesh surface (e.g., preservation of the full geometry); b) does not require normalization; c) can accommodate partial matching. Experiments conducted with public 3D models with geometric texture showcase the superiority of the mesh-LBP descriptors in comparison with competitive methods.
Frontmatter Eurographics 2015 Workshop on 3D Object Retrieval
(The Eurographics Association, 2015) Pratikakis, Ioannis; Theoharis, Theoharis; Spagnuolo, Michela; Van Gool, Luc; Veltkamp, Remco; Godil, Afzal; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Indoor Location Retrieval using Shape Matching of KinectFusion Scans to Large-Scale Indoor Point Clouds
(The Eurographics Association, 2015) Al-Nuaimi, Anas; Piccolrovazzi, Martin; Gedikli, Suat; Steinbach, Eckehard; Schroth, Georg; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
In this paper we show that indoor location retrieval can be posed as a part-in-whole matching problem of Kinect- Fusion (KinFu) query scans in large-scale target indoor point clouds. We tackle the problem with a local shape feature-based 3D Object Retrieval (3DOR) system. We specifically show that the KinFu queries suffer from artifacts stemming from the non-linear depth distortion and noise characteristics of Kinect-like sensors that are accentuated by the relative largeness of the queries. We furthermore show that proper calibration of the Kinect sensor using the CLAMS technique (Calibrating, Localizing, and Mapping, Simultaneously) proposed by Teichman et al. effectively reduces the artifacts in the generated KinFu scan and leads to a substantial retrieval performance boost. Throughout the paper we use queries and target point clouds obtained at the world's largest technical museum. The target point clouds cover floor spaces of up to 3500m2. We achieve an average localization accuracy of 6cm although the KinFu query scans make up only a tiny fraction of the target point clouds.
Morphological Analysis of 3D Faces for Weight Gain Assessment
(The Eurographics Association, 2015) Giorgi, Daniela; Pascali, Maria Antonietta; Raccichini, Giovanni; Colantonio, Sara; Salvetti, Ovidio; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
In this paper we analyse patterns in face shape variation due to weight gain. We propose the use of persistent homology descriptors to get geometric and topological information about the configuration of anthropometric 3D face landmarks. In this way, evaluating face changes boils down to comparing the descriptors computed on 3D face scans taken at different times. By applying dimensionality reduction techniques to the dissimilarity matrix of descriptors, we get a shape space in which each face is a point and face shape variations are encoded as trajectories in that space. Our first results show that persistent homology is able to identify features which are well-related to overweight, and may help assessing individual weight trends. The research is carried out in the context of the European project SEMEOTICONS, which is developing a multisensory platform which detects and monitors over time facial signs of cardio-metabolic risk.
Non-rigid 3D Shape Retrieval
(The Eurographics Association, 2015) Lian, Z.; Zhang, J.; Choi, S.; ElNaghy, H.; El-Sana, J.; Furuya, T.; Giachetti, A.; Guler, R. A.; Lai, L.; Li, C.; Li, H.; Limberger, F. A.; Martin, R.; Nakanishi, R. U.; Neto, A. P.; Nonato, L. G.; Ohbuchi, R.; Pevzner, K.; Pickup, D.; Rosin, P.; Sharf, A.; Sun, L.; Sun, X.; Tari, S.; Unal, G.; Wilson, R. C.; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Non-rigid 3D shape retrieval has become a research hotpot in communities of computer graphics, computer vision, pattern recognition, etc. In this paper, we present the results of the SHREC'15 Track: Non-rigid 3D Shape Retrieval. The aim of this track is to provide a fair and effective platform to evaluate and compare the performance of current non-rigid 3D shape retrieval methods developed by different research groups around the world. The database utilized in this track consists of 1200 3D watertight triangle meshes which are equally classified into 50 categories. All models in the same category are generated from an original 3D mesh by implementing various pose transformations. The retrieval performance of a method is evaluated using 6 commonly-used measures (i.e., PR-plot, NN, FT, ST, E-measure and DCG.). Totally, there are 37 submissions and 11 groups taking part in this track. Evaluation results and comparison analyses described in this paper not only show the bright future in researches of non-rigid 3D shape retrieval but also point out several promising research directions in this topic.
Partial 3D Object Retrieval combining Local Shape Descriptors with Global Fisher Vectors
(The Eurographics Association, 2015) Savelonas, Michalis A.; Pratikakis, Ioannis; Sfikas, Konstantinos; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
This work introduces a partial 3D object retrieval method, applicable on both meshes and point clouds, which is based on a hybrid shape matching scheme combining local shape descriptors with global Fisher vectors. The differential fast point feature histogram (dFPFH) is defined so as to extend the well-known FPFH descriptor in order to capture local geometry transitions. Local shape similarity is quantified by averaging the minimum weighted distances associated with pairs of dFPFH values calculated on the partial query and the target object. Global shape similarity is derived by means of a weighted distance of Fisher vectors. Local and global distances are derived for multiple scales and are being combined to obtain a ranked list of the most similar complete 3D objects. Experiments on the large-scale benchmark dataset for partial object retrieval of the shape retrieval contest (SHREC) 2013, as well as on the publicly available Hampson pottery dataset, support improved performance of the proposed method against seven recently evaluated retrieval methods.
Randomized Sub-Volume Partitioning for Part-Based 3D Model Retrieval
(The Eurographics Association, 2015) Furuya, Takahiko; Kurabe, Seiya; Ohbuchi, Ryutarou; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Given a query that specifies partial shape, a Part-based 3D Model Retrieval (P3DMR) system would retrieve 3D models whose part(s) matches the query. Computationally, this is quite challenging; the query must be compared against parts of 3D models having unknown position, orientation, and scale. To our knowledge, no algorithm can perform P3DMR on a database having significant size (e.g., 100K 3D models) that includes polygon soup and other not-so-well-defined shape representations. In this paper, we propose a scalable P3DMR algorithm called Part-based 3D model retrieval by Randomized Sub-Volume Partitioning, or P3D-RSVP. To match a partial query with a set of (whole) 3D models in the database, P3D-RSVP iteratively partitions a 3D model into a set of sub-volumes by using 3D grids having randomized intervals and orientations. To quickly compare the query with all the sub-volumes of all the models in the database, P3D-RSVP hashes high dimensional features into compact binary codes. Quantitative evaluation using several benchmarks shows that the P3D-RSVP is able to query a 50K model database in 2 seconds.
Range Scans based 3D Shape Retrieval
(The Eurographics Association, 2015) Godil, A.; Dutagaci, H.; Bustos, B.; Choi, S.; Dong, S.; Furuya, T.; Li, H.; Link, N.; Moriyama, A.; Meruane, R.; Ohbuchi, R.; Paulus, D.; Schreck, T.; Seib, V.; Sipiran, I.; Yin, H.; Zhang, C.; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
The objective of the SHREC'15 Range Scans based 3D Shape Retrieval track is to evaluate algorithms that match range scans of real objects to complete 3D mesh models in a target dataset. The task is to retrieve a rank list of complete 3D models that are of the same category given the range scan of a query object. This capability is essential to many computer vision systems that involves recognition and classification of objects in the environment based on depth information. In this track, the target dataset consists of 1200 3D mesh models and the query set has 180 range scans of 60 physical objects. Six research groups participated in the contest with a total of 16 different runs. This paper presents the track datasets, participants' methods and the results of the contest.
RETRIEVAL 3D: An On-line Content-Based Retrieval Performance Evaluation Tool
(The Eurographics Association, 2015) Koutsoudis, Anestis; Ioannakis, George; Pratikakis, Ioannis; Chamzas, Christos; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Performance benchmarking is an absolute necessity when attempting to objectively quantify the performance of content-based retrieval methods. For many years now, a number of plot-based and scalar-based measures in combination with benchmark datasets have already been used in order to provide objective results. In this work, we present the first version of an integrated on-line content-based retrieval evaluation tool, named RETRIEVAL 3D, which can be used in order to quantify the performance of a retrieval method. The current version of the system offers a set of popular performance measures that can be accessed through a dynamic visualisation environment. The user is able to upload retrieval results using different input data structures (e.g. binary ranked lists, floating point ranked lists, dissimilarity matrices and groundtruth data) that are already encountered in the literature including the SHREC competition series. Moreover, the system is able to provide evaluation mechanisms for known within the retrieval research community benchmark datasets. It offers performance measures parameterisation that enables the user to determine specific aspects of the evaluated retrieval method. Performance reports archiving and downloading are some of the system's user-oriented functionalities.
Retrieval of Non-rigid (textured) Shapes Using Low Quality 3D Models
(The Eurographics Association, 2015) Giachetti, Andrea; Farina, Francesco; Fornasa, Francesco; Tatsuma, Atsushi; Sanada, Chika; Aono, Masaki; Biasotti, Silvia; Cerri, Andrea; Choi, Sungbin; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
This paper reports the results of the SHREC 2015 track on retrieval of non-rigid (textured) shapes from low quality 3D models. This track has been organized to test the ability of the algorithms recently proposed by researchers for the retrieval of articulated and textured shapes to deal with real-world deformations and acquisition noise. For this reason we acquired with low cost devices models of plush toys lying on different sides on a platform, with articulated deformations and with different illumination conditions. We obtained in this way three novel and challenging datasets that have been used to organize a contest where the proposed task was the retrieval of istances of the same toy within acquired shapes collections, given a query model. The differences in datasets and tasks were related to the fact that one dataset was built without applying texture to shapes, and the others had texture applied to vertices with two different methods. We evaluated the retrieval results of the proposed techniques using standard evaluation measures: Precision-Recall curve; E-Measure; Discounted Cumulative Gain; Nearest Neighbor, First- Tier (Tier1) and Second-Tier (Tier2), Mean Average Precision. Robustness of methods against texture and shape deformation has also been separately evaluated.
Retrieval of Objects Captured with Kinect One Camera
(The Eurographics Association, 2015) Pascoal, Pedro B.; Proença, Pedro; Gaspar, Filipe; Dias, Miguel Sales; Teixeira, Filipe; Ferreira, Alfredo; Seib, Viktor; Link, Norman; Paulus, Dietrich; Tatsuma, Atsushi; Aono, Masaki; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Low-cost RGB-D sensing technology, such as the Microsoft Kinect, is gaining acceptance in the scientific community and even entering into our homes. This technology enables ordinary users to capture everyday object into digital 3D representations. Considering the image retrieval context, whereas the ability to digitalize photos led to a rapid increase of large collections of images, which in turn raised the need of efficient search and retrieval techniques. We believe the same is happening now for the 3D domain. Therefore, it is essential to identify which 3D shape descriptors, provide better matching and retrieval of such digitalized objects. In this paper, we start by presenting a collection of 3D objects acquired using the latest version of Microsoft Kinect, namely, Kinect One. This dataset, comprising 175 common household objects classified into 18 different classes, was then used for the SHape REtrieval Contest (SHREC). Two groups have submitted their 3D matching techniques, providing the rank list with top 10 results, using the complete set of 175 objects as queries.
Scalability of Non-Rigid 3D Shape Retrieval
(The Eurographics Association, 2015) Sipiran, I.; Bustos, B.; Schreck, T.; Bronstein, A. M.; Bronstein, M.; Castellani, U.; Choi, S.; Lai, L.; Li, H.; Litman, R.; Sun, L.; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Due to recent advances in 3D acquisition and modeling, increasingly large amounts of 3D shape data become available in many application domains. This rises not only the need for effective methods for 3D shape retrieval, but also efficient retrieval and robust implementations. Previous 3D retrieval challenges have mainly considered data sets in the range of a few thousands of queries. In the 2015 SHREC track on Scalability of 3D Shape Retrieval we provide a benchmark with more than 96 thousand shapes. The data set is based on a non-rigid retrieval benchmark enhanced by other existing shape benchmarks. From the baseline models, a large set of partial objects were automatically created by simulating a range-image acquisition process. Four teams have participated in the track, with most methods providing very good to near-perfect retrieval results, and one less complex baseline method providing fair performance. Timing results indicate that three of the methods including the latter baseline one provide near- interactive time query execution. Generally, the cost of data pre-processing varies depending on the method.
Sketch-based 3D Object Retrieval Using Two Views and a Visual Part Alignment
(The Eurographics Association, 2015) Yasseen, Zahraa; Verroust-Blondet, Anne; Nasri, Ahmad; I. Pratikakis and M. Spagnuolo and T. Theoharis and L. Van Gool and R. Veltkamp
Hand drawn figures are the imprints of shapes in human's mind. How a human expresses a shape is a consequence of how he or she visualizes it. A query-by-sketch 3D object retrieval application is closely tied to this concept from two aspects. First, describing sketches must involve elements in a figure that matter most to a human. Second, the representative 2D projection of the target 3D objects must be limited to ''the canonical views'' from a human cognition perspective. We advocate for these two rules by presenting a new approach for sketch-based 3D object retrieval that describes a 2D shape by the visual protruding parts of its silhouette. Furthermore, the proposed approach computes estimations of ''part occlusion'' and ''symmetry'' in 2D shapes in a new paradigm for viewpoint selection that represents 3D objects by only the two views corresponding to the minimum value of each.

Browse

Browsing 3DOR 15 by Title

Results Per Page

Sort Options