TPTomislav Petković

Tomislav Petković

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Tomislav Petković has a total of 31 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Phase-Shifted Quasi-Elliptical Fringe Patterns for Underwater 3-D Imaging

    by Domagoj Zoraja, Tomislav Petković, Josep Forest

    The development of robust fringe patterns for an underwater imaging is challenging due to the unique physical properties of the underwater environment. A line in the image plane, when projected through a flat refractive interface, deforms which introduces errors in both calibration and reconstruction procedures. In this paper, we present a novel image formation model, based on the axial camera model, which utilizes phase shifted quasi-elliptical fringe patterns for the underwater imaging. When projected through flat refractive interfaces, deformed elliptical fringe patterns produce concentric cones of camera rays in the outmost medium, each uniquely identified by its radius encoded in the phase. Triangulation is then achieved by computing ray-cone intersections, with extraneous solutions filtered out since more than one viable solutions exist. The proposed quasi-elliptical fringe pattern was experimentally evaluated by imaging calibration objects and performing 3D reconstruction, wherein quantitative results demonstrate the feasibility of the proposed method.

  • Pose-Independent 3D Anthropometry from Sparse Data

    by David Bojanić, Stefanie Wuhrer, Tomislav Petković, Tomislav Pribanić


  • Open Access

    Pose-independent 3D Anthropometry from Sparse Data

    by David Bojanić, Stefanie Wuhrer, Tomislav Petković, Tomislav Pribanić

    3D digital anthropometry is the study of estimating human body measurements from 3D scans. Precise body measurements are important health indicators in the medical industry, and guiding factors in the fashion, ergonomic and entertainment industries. The measuring protocol consists of scanning the whole subject in the static A-pose, which is maintained without breathing or movement during the scanning process. However, the A-pose is not easy to maintain during the whole scanning process, which can last even up to a couple of minutes. This constraint affects the final quality of the scan, which in turn affects the accuracy of the estimated body measurements obtained from methods that rely on dense geometric data. Additionally, this constraint makes it impossible to develop a digital anthropometry method for subjects unable to assume the A-pose, such as those with injuries or disabilities. We propose a method that can obtain body measurements from sparse landmarks acquired in any pose. We make use of the sparse landmarks of the posed subject to create pose-independent features, and train a network to predict the body measurements as taken from the standard A-pose. We show that our method achieves comparable results to competing methods that use dense geometry in the standard A-pose, but has the capability of estimating the body measurements from any pose using sparse landmarks only. Finally, we address the lack of open-source 3D anthropometry methods by making our method available to the research community at https://github.com/DavidBoja/pose-independent-anthropometry.

  • Open Access

    Neural Network Meta Classifier: Improving the Reliability of Anomaly Segmentation

    by Jurica Runtas, Tomislav Petkovic

    Deep neural networks (DNNs) are a contemporary solution for semantic segmentation and are usually trained to operate on a predefined closed set of classes. In open-set environments, it is possible to encounter semantically unknown objects or anomalies. Road driving is an example of such an environment in which, from a safety standpoint, it is important to ensure that a DNN indicates it is operating outside of its learned semantic domain. One possible approach to anomaly segmentation is entropy maximization, which is paired with a logistic regression based post-processing step called meta classification, which is in turn used to improve the reliability of detection of anomalous pixels. We propose to substitute the logistic regression meta classifier with a more expressive lightweight fully connected neural network. We analyze advantages and drawbacks of the proposed neural network meta classifier and demonstrate its better performance over logistic regression. We also introduce the concept of informative out-of-distribution examples which we show to improve training results when using entropy maximization in practice. Finally, we discuss the loss of interpretability and show that the behavior of logistic regression and neural network is strongly correlated.

  • Open Access

    Addressing the generalization of 3D registration methods with a featureless baseline and an unbiased benchmark

    by David Bojanić, Kristijan Bartol, Josep Forest, Tomislav Petković, Tomislav Pribanić

    Recent 3D registration methods are mostly learning-based that either find correspondences in feature space and match them, or directly estimate the registration transformation from the given point cloud features. Therefore, these feature-based methods have difficulties with generalizing onto point clouds that differ substantially from their training data. This issue is not so apparent because of the problematic benchmark definitions that cannot provide any in-depth analysis and contain a bias toward similar data. Therefore, we propose a methodology to create a 3D registration benchmark, given a point cloud dataset, that provides a more informative evaluation of a method w.r.t. other benchmarks. Using this methodology, we create a novel FAUST-partial (FP) benchmark, based on the FAUST dataset, with several difficulty levels. The FP benchmark addresses the limitations of the current benchmarks: lack of data and parameter range variability, and allows to evaluate the strengths and weaknesses of a 3D registration method w.r.t. a single registration parameter. Using the new FP benchmark, we provide a thorough analysis of the current state-of-the-art methods and observe that the current method still struggle to generalize onto severely different out-of-sample data. Therefore, we propose a simple featureless traditional 3D registration baseline method based on the weighted cross-correlation between two given point clouds. Our method achieves strong results on current benchmarking datasets, outperforming most deep learning methods. Our source code is available on github.com/DavidBoja/exhaustive-grid-search.


  • Open Access

    Calibration of a Structured Light Imaging System in Two-Layer Flat Refractive Geometry for Underwater Imaging

    by Domagoj Zoraja, Tomislav Petković, Josep Forest, Tomislav Pribanić

    The development of a robust 3D imaging system for underwater applications is a crucial process in underwater imaging where the physical properties of the underwater environment make the implementation of such systems challenging. Calibration is an essential step in the application of such imaging systems and is performed to acquire the parameters of the image formation model and to enable 3D reconstruction. We present a novel calibration method for an underwater 3D imaging system comprising a pair of cameras, of a projector, and of a single glass interface that is shared between cameras and projector(s). The image formation model is based on the axial camera model. The proposed calibration uses a numerical optimization of a 3D cost function to determine all system parameters, thus avoiding the minimization of re-projection errors which require numerically solving a 12th order polynomial equation multiple times for each observed point. We also propose a novel stable approach to estimate the axis of the axial camera model. The proposed calibration was experimentally evaluated on four different glass interfaces, wherein several quantitative results were reported, including the re-projection error. The achieved mean angular error of the system’s axis was under 6∘, and the mean absolute errors for the reconstruction of a flat surface were 1.38 mm for normal glass interfaces and 2.82 mm for the laminated glass interface, which is more than sufficient for application.


  • Open Access

    Multiprojector Multicamera Structured Light Surface Scanner

    by Tomislav Petkovic, Tomislav Pribanic

    Structured light surface scanners are usually limited to a single projector and cannot provide a full surface profile in a single scan. Multiple projectors must be employed to extend the imaging area over the whole surface of the imaged object. Using multiple projectors is a challenging problem due to inter-projector interference which makes pattern decoding hard. We describe a practical approach to multi-projector structured light surface scanning based on sinusoidal fringe patterns and temporal phase shifting which allows simultaneous projector employment. In our approach a different set of temporal phase shifts is assigned to each projector to enable an efficient and robust separation of observed combined patterns into the contributions of each projector. The proposed approach does not place any restrictions on projector placement nor on the upper limit of the number of projectors used. We demonstrate the applicability of the proposed approach on a structured light surface scanner comprised of three projectors and six cameras.


  • Open Access

    Linear Regression vs. Deep Learning: A Simple Yet Effective Baseline for Human Body Measurement

    by Kristijan Bartol, David Bojanić, Tomislav Petković, Stanislav Peharec, Tomislav Pribanić

    We propose a linear regression model for the estimation of human body measurements. The input to the model only consists of the information that a person can self-estimate, such as height and weight. We evaluate our model against the state-of-the-art approaches for body measurement from point clouds and images, demonstrate the comparable performance with the best methods, and even outperform several deep learning models on public datasets. The simplicity of the proposed regression model makes it perfectly suitable as a baseline in addition to the convenience for applications such as the virtual try-on. To improve the repeatability of the results of our baseline and the competing methods, we provide guidelines toward standardized body measurement estimation.


  • Open Access

    A Review of Body Measurement Using 3D Scanning

    by Kristijan Bartol, David Bojanic, Tomislav Petkovic, Tomislav Pribanic

    The understanding of body measurements and body shapes in and between populations is important and has many applications in medicine, surveying, the fashion industry, fitness, and entertainment. Body measurement using 3D surface scanning technologies is faster and more convenient than measurement with more traditional methods and at the same time provides much more data, which requires automatic processing. A multitude of 3D scanning methods and processing pipelines have been described in the literature, and the advent of deep learning-based processing methods has generated an increased interest in the topic. Also, over the last decade, larger public 3D human scanning datasets have been released. This paper gives a comprehensive survey of body measurement techniques, with an emphasis on 3D scanning technologies and automatic data processing pipelines. An introduction to the three most common 3D scanning technologies for body measurement, passive stereo, structured light, and time-of-flight, is provided, and their merits w.r.t. body measurement are discussed. Methods described in the literature are discussed within the newly proposed framework of five common processing stages: preparation, scanning, feature extraction, model fitting, and measurement extraction. Synthesizing the analyzed prior works, recommendations on specific 3D body scanning technologies and the accompanying processing pipelines for the most common applications are given. Finally, an overview of about 80 currently available 3D scanners manufactured by about 50 companies, as well as their taxonomy regarding several key characteristics, is provided in the Appendix.