WSCG 2019

27. International Conference on
Computer Graphics, Visualization and Computer Vision 2019

Primavera Hotel and Congress Center
Plzen, Czech Republic

May 27 – 30, 2019







Preliminary Program

Schedule within a day can change without a notice
There is only access to the abstracts papers, now
PDF of papers will be available during the conference




Conference Chair

Prof. Vaclav Skala
University of West Bohemia
Plzen, Czech Republic





Registration – Primavera Hotel and Congress Center

·        Monday 17:00 – 20:00

·        Every conference day during breaks


Technical Sessions

·        Tuesday – Wednesday – all the days

·        Thursday – morning
afternoon Pilsen city short guided tour & beer tasting
Details at:


Conference Dinner

·        To be announced, preliminary scheduled to Wednesday
check the Information/Registration desk of the WSCG conference


Supporting organizations

·        Department of Computer Science and Enginnering, Faculty of Applied Sciences
University of West Bohemia, Plzen, Czech Republic

·        NVIDIA, Czech Republic,

·        Center of Computer Graphics and Visualization, Univ. of West Bohemia

WSCG 2019





Session A Tuesday 8:30 - 10:10

·          Besenthal, S., Maisch, S., Ropinski, T.: Multi-Resolution Rendering for Computationally Expensive Lighting Effects
Paper Code: B17 (Full)
Abstract Paper Additional File

·          Kobrtek, J., Milet, T., Herout, A.: Silhouette Extraction for Shadow Volumes Using Potentially Visible Sets
Paper Code: A41 (Full)
Abstract Paper Additional File

·          Listemann, M., Trapp, M., Döllner, J.: Lens-based Focus+Context Visualization Techniques for Interactive Exploration of Web-based Reachability Maps
Paper Code: C17 (Full)
Abstract Paper

·          Sugihara,K.: Straight Skeleton Computation Optimized for Roof Model Generation
Paper Code: C67 (Full)
Abstract Paper Additional File

·          Shcherbakov,A.,Frolov,V.: Dynamic Radiosity
Paper Code: C47 (Full)
Abstract Paper


Session B Tuesday 10:30 - 12:10

·          Schult,T., Klose,U., Hauser,T.-K., Ehricke,H.-H.: Anatomy-Focused Volume Line Integral Convolution for Brain White Matter Visualization
Paper Code: C41 (Full)
Abstract Paper

·          Ogniewski, J.: Cubic Spline Interpolation in Real-Time Applications using Three Control Points
Paper Code: A43 (Full)
Abstract Paper

·          Merten,N., Saalfeld,S., Preim,B.: Floor Map Visualizations of Medical Volume Data
Paper Code: B05 (Full)
Abstract Paper Additional File

·          Brunet J-N, Magnoux V., Ozell B., Cotin S.: Corotated meshless implicit dynamics for deformable bodies
Paper Code: C53 (Full)
Abstract Paper

·          Belyaev, S.Y., Smirnov, P.O., Smirnova, N.D., Shubnikov, V.G.: Fast Adaptive Undersampling for Volume Rendering
Paper Code: A37 (Full)
Abstract Paper Additional File


Session C - Welcome - Tuesday 12:10 - 12:30




Session D Tuesday 14:00 - 15:40

·          Qi,Z., Lee,J.K.: Automated Object Tracking in Sterile Pharmacy Compounding
Paper Code: D19 (Full)
Abstract Paper Additional File

·          Lykke,J.R., Olsen,A.B., Berman,P., Bærentzen,J.A., Frisvad,J.R.: Accounting for Object Weight in Interaction Design for Virtual Reality
Paper Code: D07 (Full)
Abstract Paper Additional File

·          Khaksari Haddad,H., Laurendeau,D.: Alignment of Point Clouds for Comparison of Infrastructures in Civil Engineering on Quality Control in Metrology
Paper Code: D59 (Full)
Abstract Paper

·          Clinton L. Jeffery: The City Metaphor in Software Visualization
Paper Code: D47 (Full)
Abstract Paper Additional File

·          Lefkovits, S., Lefkovits, L., Szilágyi, L.: CNN Approaches for Dorsal Hand Vein Based Identification
Paper Code: C59 (Short)
Abstract Paper Additional File


Session E Tuesday 16:00 - 18:00

·          Hauenstein, J., Newman, T.: Exhibition and Evaluation of Two Schemes for Determining Hypersurface Curvature in Volumetric Data
Paper Code: D02 (Full)
Abstract Paper Additional File

·          Sekmen, S., Akyüz, A. O.: Compressed Exposure Sequences for HDR Imaging
Paper Code: D43 (Full)
Abstract Paper Additional File

·          Raposo,A.N., Gomes,A.J.P.: Pi-surfaces: products of implicit surfaces towards constructive composition of 3D objects
Paper Code: C83 (Full)
Abstract Paper Additional File

·          Gorbachev,V., Kaynarova,E., Makarov,A., Yakovleva,E.: Grayscale and Color Basis Images
Paper Code: D03 (Full)
Abstract Paper Additional File

·          Bustacara-Medina, C., Flórez-Valencia, L.: An automatic stopping criterion for nonlinear anisotropic diffusion
Paper Code: A89 (Full)
Abstract Paper

·          Daoud,Z., Ben Hamida, A., & Ben Amar,C: Automatic video fire detection approach based on PJF color modeling and spatio-temporal analysis
Paper Code: A73 (Full)
Abstract Paper Additional File

·          Basov,A.Y., Budak,V.P., Grimailo,A.V.: The Role of Polarization in The Multiple Reflections Modeling
Paper Code: B73 (Poster) Abstract Paper Additional File


Session F Wednesday 8:30 - 10:10

·          Organisciak,D., Riachy,C., Aslam,N., Shum, H.P.H.: Triplet Loss with Channel Attention for Person Re-identification
Paper Code: D61 (Full)
Abstract Paper Additional File

·          Lee,H., Sa,J., Shin,H., Chung,Y., Park,D., Kim,H.: Deep Learning-based Overlapping-Pigs Separation by Balancing Accuracy and Execution Time
Paper Code: B03 (Full)
Abstract Paper

·          Hausdorf,A., Hinzmann,N., Zeckzer,D.: SyCaT-Vis: Visualization-Based Support of Analyzing System Behavior based on System Call Traces
Paper Code: C11 (Full)
Abstract Paper

·          Riachy,C., Al-Maadeed,N., Organisciak,D., Khelifi,F., Bouridane,A.: 3D Gaussian Descriptor for Video-based Person Re-Identification
Paper Code: D67 (Full)
Abstract Paper Additional File

·          Möller,J., Meyer,B., Eisemann,M.: Porting A Visual Inertial SLAM Algorithm To Android Devices
Paper Code: A79 (Short)
Abstract Paper


Session G Wednesday 10:30 - 12:10

·          Leipnitz,A., Strutz,T., Jokisch, O.: Performance Assessment of Convolutional Neural Networks for Semantic Image Segmentation
Paper Code: B61 (Full)
Abstract Paper Additional File

·          Roig-Maimo, MF., Mas-Sanso,R.: Collateral effects of the Kalman Filter on the Throughput of a Head-Tracker for Mobile Devices
Paper Code: C89 (Full)
Abstract Paper Additional File

·          Qureshi,H.S., Wizcorek,R.: Curb Detection for a Pedestrian Assistance System using End-to-End Learning
Paper Code: C02 (Full)
Abstract Paper Additional File

·          Munoz-Arango,Juan S., Reiners,D., Cruz-Neira, C.: Analyzing pixel spread correlation with lenticular lens efficiency on multi user VR displays
Paper Code: C19 (Full)
Abstract Paper Additional File

·          Khemmar,R., Gouveai,M., Decoux,D., Ertaud,J.Y.: Real Time Pedestrian and Object Detection and Tracking-based Deep Learning. Application to Drone Visual Tracking
Paper Code: B83 (Short)
Abstract Paper Additional File


Session H - Common Photo - Wednesday 12:10 - 12:30



Session K Wednesday 14:00 - 15:40

·          Shirley, T., Presnov, D., Kolb, A.: A Lightweight Approach to 3D Measurement of Chronic Wounds
Paper Code: B67 (Full)
Abstract Paper Additional File

·          Wallelign,S., Polceanu M., Jemal T., Buche C.,: Coffee Grading with Convolutional Neural Networks using Small Datasets with High Variance
Paper Code: C61 (Full)
Abstract Paper Additional File

·          Rotger, G., Moreno-Noguer, F., Lumbreras, F., Agudo, A.: Detailed 3D Face Reconstruction from a Single RGB Image
Paper Code: C43 (Full)
Abstract Paper Additional File

·          Wegen, O., Trapp, M., Döllner, J., S. Pasewaldt: Performance Evaluation and Comparison of Service-based Image Processing based on Software Rendering
Paper Code: D05 (Full)
Abstract Paper

·          Bouressace Hassina: Title Segmentation in Arabic Document Pages
Paper Code: B89 (Short)
Abstract Paper


Session L Wednesday 16:00 - 17:00

·          Smolik,M.: Radial Basis Functions and Space Subdivision for Large Scattered Vector Fiields Approximation
Paper Code: Z1 (Tutorial)
Abstract Paper Additional File

·          Katragadda,S.: Identification of stereo rig alignment error based on vertical disparity map
Paper Code: D17 (Full)
Abstract Paper Additional File

·          Barina,D., Chlubna,T., Solony,M., Dlabaja,D., Zemcik,P.: Evaluation of 4D Light Field Compression Methods
Paper Code: C13 (Full)
Abstract Paper Additional File

·          Zechmeister, S., Cornel, D., Waser, J.: 3D Annotations for Geospatial Decision Support Systems
Paper Code: D13 (Full)
Abstract Paper Additional File

·          Hemani M., Sinha A., Balaji K.: Stylized Sketch Generation using Convolutional Networks
Paper Code: B97 (Full)
Abstract Paper Additional File

·          Pérez Gutierrez, M., Córdova-Cruzatty, A.: Obstacle detection algorithm by means of images with a ZED camera using ROS software in a drone
Paper Code: D29 (Short)
Abstract Paper








Is to be organized outside

Order DISCOUNTED ticket for the Conference Dinner ticket at the
Conference REGISTRATION desk – places are limited
(Monday or Tuesday morning)

!!! DINNER IS NOT at the Primavera Hotel and Congress Center !!!


Session M Thursday 8:30 - 10:10

·          Friedrich,M., Guimerà Cuevas,F., Sedlmeier,A., Ebert,A.: Evolutionary Generation of Primitive-Based Mesh Abstractions
Paper Code: A61 (Full)
Abstract Paper Additional File

·          Novikov,M. M., Galeev,R. M., Knyaz,V. A.,: Creating digital models of paleontological sample by photogrammetry and computed tomography
Paper Code: A71 (Full)
Abstract Paper Additional File

·          Cibulski,L.,May,T.,Preim,B.,Bernard,J.,Kohlhammer,J.: Visualizing Time Series Consistency for Feature Selection
Paper Code: C37 (Full)
Abstract Paper Additional File

·          Marín,C., Chover,M., Sotoca,J.M.: Prototyping a game engine architecture as a multi-agent system
Paper Code: B29 (Short)
Abstract Paper Additional File


Session N Thursday 10:30 - 12:00

·          Karolyi,M., Krejci,J., Štavnický,J., Vyškovský,R., Komenda,M.: Tools for development of interactive web-based maps: application in healthcare
Paper Code: A31 (Short)
Abstract Paper

·          Blum,S., Cetin,G., Stuerzlinger,W.: Immersive Analytics Sensemaking on Different Platforms
Paper Code: D31 (Short)
Abstract Paper Additional File

·          Saddiqa, M., Rasmussen, L., Larsen, B., Magnussen, R., Pedersen, J.M.: Open Data Visualization in Danish Schools: A case Study
Paper Code: B07 (Short)
Abstract Paper Additional File

·          Silva,M., Martino,J.: Improving facial attraction in videos
Paper Code: D23 (Short)
Abstract Paper

·          Trapp, M., Pasewaldt, S., Döllner, J.: Techniques for GPU-based Color Palette Mapping
Paper Code: D71 (Short)
Abstract Paper

·          Dubrovskaya, M: Automatic Human Heart Search on CT-Scans
Paper Code: B43 (Short)
Abstract Paper


Session P - Closing Session - Thursday 12:00 - 12:10





Guided tour on Thursday will probably start at 15:00 at the Center of the Plzen City
Information will be on the conference registration desk.



INFORMATION for Presenters

See the WEB site for latest details


It is recommended to use your own computer for presentation.


Presenters are requested to verify presentation in advance – time slot is to be kept strictly


·        Data projector with standard PC interface

·        PC (not fast) with Windows


If you use Apple or other platform, make sure that you have all the convertors for power supply and data projector connection









A37: Fast Adaptive Undersampling for Volume Rendering

Belyaev, S.Y., Smirnov, P.O., Smirnova, N.D., Shubnikov, V.G.

Adaptive undersampling is a method for accelerating the rendering process by replacing the calculation of a volume integral with an interpolation procedure for a number of pixels. In this paper, we propose a method for accelerating the volume integral calculation for the rest of the pixels, i.e. those pixels for which interpolation cannot be done with sufficient accuracy. This method requires two passes through the input data. On the first pass, rendering is done into a low-resolution texture. At this stage, the values of the volume integral on a set of intervals of a given length are calculated and saved into a special G-buffer alone with the pixel’s color. On the second pass, these values are used to determine colors of the pixels. For those pixels whose result is not precise enough, the volume integral is calculated on one or several intervals, rather than the whole ray. The proposed method allows one to accelerate adaptive undersampling by a factor of 1.5 on average, depending on the input data.


A41: Silhouette Extraction for Shadow Volumes Using Potentially Visible Sets

Kobrtek, J., Milet, T., Herout, A.

In this paper we present a novel approach for accelerated silhouette computation based on potentially visible sets stored in the octree acceleration structure. Scene space, where the light source can appear, is subdivided into voxels. The octree voxels contain two precomputed sets of edges that potentially or always belong to the silhouette. We also propose a novel method of octree compression for reduction of the memory footprint of the resulting acceleration structure. Using our novel technique we were able to considerably decrease the silhouette computational complexity and reduce its sensitivity to the number of edges.


D19: Automated Object Tracking in Sterile Pharmacy Compounding

Qi,Z., Lee,J.K.

A new method to automatically track objects in the process of sterile pharmacy compounding is introduced. The method performs the automatic object tracking by applying a simple object detection followed by an object tracking algorithm. 
The advantages of the method are: it is simple and efficient; it enables both single and multiple objects tracking; and it does not need large sets of templates, nor any training set. The experimental results show that the method reasonably tracks the objects accurately.


A43: Cubic Spline Interpolation in Real-Time Applications using Three Control Points

Ogniewski, J.

Spline interpolation is widely used in many different applications like computer graphics, animations and robotics. Many of these applications are run in real-time with constraints on computational complexity, thus fueling the need for computational inexpensive, real-time, continuous and loop-free data interpolation techniques.
Often Catmull-Rom splines are used, which use four control points: the two points between which to interpolate as well as the point directly before and the one directly after. If interpolating over time, this last point will lie in the future. However, in real-time applications future values may not be known in advance, meaning that Catmull-Rom splines are not applicable.
In this paper we introduce another family of interpolation splines (dubbed Three-Point-Splines) which show the same characteristics as Catmull-Rom, but which use only three control points, omitting the one “in the future”. Therefore they can generate smooth interpolation curves even in applications which do not have knowledge of future points, without the need for more computational complex methods. The generated curves are more rigid than Catmull-Rom, and because of that the Three-Point-Splines will not generate self-intersections within an interpolated curve segment, a property that has to be introduced to Catmull-Rom by careful parameterization. Thus, the Three-Point-Splines allow for greater freedom in parameterization, and can therefore be adapted to the application at hand, e.g. to a requested curvature or limitations on acceleration/deceleration.
We will also show a method that allows to change the control points during an ongoing interpolation, both with Thee-Point-Splines as well as with Catmull-Rom splines.

 A61: Evolutionary Generation of Primitive-Based Mesh Abstractions

Friedrich,M., Guimerà Cuevas,F., Sedlmeier,A., Ebert,A.

The procedural generation of data sets for empirical algorithm validation and deep learning tasks in the area of primitive-based geometry is cumbersome and time-consuming while ready-to-use data sets are rare. 
We propose a new and highly flexible framework based on Evolutionary Computing that is able to create primitive-based abstractions of existing triangle meshes favoring fast running times and high geometric variation over reconstruction precision.
These abstractions are represented as CSG trees to widen the scope of possible applications. 
As part of the evaluation, we show how we successfully used the generator to create a data set for the evaluation of neural point cloud segmentation pipelines and additionally explain how to use the system to create artistic abstractions of meshes provided by publicly available triangle mesh databases.


A71: Creating digital models of paleontological sample by photogrammetry and computed tomography

Novikov,M. M., Galeev,R. M., Knyaz,V. A.,

The specificity of paleoanthropological research and accurate anthropological documentation is due to the complexity of the form of anthropological objects, their uniqueness, high historical and scientific value. The main purpose of this work is to develop methods for creating digital 3D models with a high degree of informativity for the development of digital documentation systems of paleoanthropological objects. The article presents a comparative analysis of digital models of anthropological objects obtained using photogrammetry and computed tomography. It is shown that the combined use of non-contact methods of videogrammetry and computed tomography allows to create high-precision three-dimensional models with photorealistic texture and accurate internal and hidden geometry. The proposed approach allows not only to create virtual collections for wide sharing of specialists, but also using modern methods of additive manufacturing to make exact copies of unique artifacts.


A73: Automatic video fire detection approach based on PJF color modeling and spatio-temporal analysis

Daoud,Z., Ben Hamida, A., & Ben Amar,C

Recently, due to the huge damage caused by fires in many countries in the world, fire detection is getting more and more interest as an increasing important issue. Nowadays, the early fire detection in video surveillance scenes is emerging as an alternative solution to overcome the shortcomings of the current inefficient sensors. In this paper, we propose a new video based-fire detection method exploiting color and motion information of fire. Our approach consists in detecting all moving regions in the scene to select then areas likely to be fire. Further, motion analysis is required to identify the accurate fire regions. The proposed method is evaluated on different video datasets containing diverse fire and non-fire videos. Experimental results demonstrate the effectiveness of our proposed method by achieving high fire detection and low false alarms rates. Moreover, it greatly outperforms the related works with 98.81 % accuracy and only 2 % of false positive rate.


A89: An automatic stopping criterion for nonlinear anisotropic diffusion

Bustacara-Medina, C., Flórez-Valencia, L.

Nonlinear anisotropic diffusion (NAD) filtering is a procedure based on nonlinear evolution PDEs which seeks to improve images qualitatively by removing noise while preserving details and even enhancing edges. However, well-known implementations are sensitive to parameters which are necessarily tuned to sharpen a narrow range of edge slopes; otherwise, edges are either blurred or staircased. One important parameter is the iterations number, for that reason, in this paper is proposed a stopping criterion for the diffusion process. To meet this goal, two stopping criteria were compared. The first is the stopping criterion proposed by Joao et. al., which is based on the Mean Squared Error (MSE). The second is our proposed method based on the CIRR contrast measure. To this end, a comparative analysis of five diffusion methods is performed. Four of them are nonlinear anisotropic diffusion methods and the fifth is the Perona-Malik method. According to the tests performed, the number of iterations required by the smoothing algorithms using the proposed stopping criterion is lower.


B03: Deep Learning-based Overlapping-Pigs Separation by Balancing Accuracy and Execution Time

Lee,H., Sa,J., Shin,H., Chung,Y., Park,D., Kim,H.

The crowded environment of a pig farm is highly vulnerable to the spread of infectious diseases such as foot-and-mouth disease, and studies have been conducted to automatically analyze behavior of pigs in a crowded pig farm through a video surveillance system using a top-view camera. Although it is required to correctly separate overlapping-pigs for tracking each individual pigs, extracting the boundaries of each pig fast and accurately is a challenging issue due to the complicated occlusion patterns such as X shape and T shape. In this study, we propose a fast and accurate method to separate overlapping-pigs not only by exploiting the advantage (i.e., one of the fast deep learning-based object detectors) of You Only Look Once, YOLO, but also by overcoming the disadvantage (i.e., the axis aligned bounding box-based object detector) of YOLO with the test-time data augmentation of rotation. Experimental results with the occlusion patterns between the overlapping-pigs show that the proposed method can provide better accuracy and faster processing speed than one of the state-of-the-art deep learning-based segmentation techniques such as Mask R-CNN (i.e., the performance improvement over Mask R-CNN was about 11 times, in terms of the accuracy/processing speed performance metrics).


B05: Floor Map Visualizations of Medical Volume Data

Merten,N., Saalfeld,S., Preim,B.

Typically, volumetric medical image data is examined by assessing each slice of an image stack individually. However, this enables observers to assess in-plane spatial relationships between anatomical structures only and requires them to keep track of relationships along the third anatomical plane mentally. Therefore, visualization techniques are researched to support this task by depicting spatial information along the third plane, but they can introduce a high degree of abstraction. To overcome this, we present a novel approach that transforms image stacks with labeled anatomical structures into maps with a three-dimensional layout, namely floor maps. Since this approach increases the visual complexity under certain conditions, some clinical application scenarios, e.g. diagnosis and therapy planning, probably will not benefit. Thus, the approach is mainly aimed to support student training and the generation of clinical reports. We also discuss how to enhance the slice-based exploration of medical image stacks via floor maps and present the results of an informal evaluation with three trained anatomists.


B17: Multi-Resolution Rendering for Computationally Expensive Lighting Effects

Besenthal, S., Maisch, S., Ropinski, T.

Many lighting methods used in computer graphics such as indirect illumination can have very high computational costs and need to be approximated for real time applications. These costs can be reduced significantly by using upsampling techniques which tend to introduce artifacts and affect the visual quality of the rendered image. This paper introduces a versatile approach for accelerating the rendering of screen space methods while maintaining the visual quality. This is achieved by exploiting the low frequency nature of many of these illumination methods and the geometrical continuity of the scene. First the screen space is dynamically divided into separate subimages, then the lighting is rendered for each subimage in an adequate resolution and lastly the subimages are put together in order to compose the final image. This is done by identifying edges in the scene and generating masks precisely specifying which part of the image is included in which subimage. The masks therefore determine which part of the image is rendered in which resolution. The upsampling and merging process of the subimages is done stepwise and allows optically soft transitions between the different resolution levels. For this paper the introduced multi-resolution rendering method was implemented and tested on three commonly used lighting methods. These are screen space ambient occlusion, soft shadow mapping and screen space global illumination.


B61: Performance Assessment of Convolutional Neural Networks for Semantic Image Segmentation

Leipnitz,A., Strutz,T., Jokisch, O.

Convolutional neural networks are applied successfully for image classification and object detection. Recently, they have been adopted to semantic segmentation tasks and several new network architectures have been proposed. With respect to automotive applications, the Cityscapes dataset is often used as a benchmark. It is one of the biggest datasets in this field and consists of a training, a validation, and a test set. While training and validation allow the optimisation of these nets, the test dataset can be used to evaluate their performance.
Our investigations have shown that while these networks perform well for images of the Cityscapes dataset, their segmentation quality significantly drops when applied to new data. It seems that they have limited generalisation abilities. In order to find out whether the image content itself or other image properties cause this effect, we have carried out systematic investigations with modified Cityscapes data. We have found that camera-dependent image properties like brightness, contrast, or saturation can significantly influence the segmentation quality. This papers presents the results of these tests including eight state-of-the-art CNNs. It can be concluded that the out-of-the-box usage of CNNs in real-world environments is not recommended.


B67: A Lightweight Approach to 3D Measurement of Chronic Wounds

Shirley, T., Presnov, D., Kolb, A.

This paper presents a light-weight process for 3D reconstruction and measurement of chronic wounds using a commonly available smartphone as an image capturing device. The first stage of our measurement pipeline comprises the creation of a dense 3D point cloud using structure-from-motion (SfM). Furthermore, the wound area is segmented from the surrounding skin using dynamic thresholding in CIELAB color space and a surface is estimated to simulate the missing skin in the
wound area. Together with a mesh reconstruction of the wound, the skin surface and the segmented wound is used to calculate the wound dimensions, i.e., its length, surface area and volume. We evaluate the presented pipeline using three wound phantoms, representing different stages in healing, and compare the subsequently scanned and measured wound dimensions with manually measured ones.


B97: Stylized Sketch Generation using Convolutional Networks

Hemani M., Sinha A., Balaji K.

The task of synthesizing sketches from photographs has been pursued with image processing methods and supervised learning based approaches. The former lack flexibility and the latter require large quantities of ground-truth data which is hard to obtain because of the manual effort required. We present a convolutional neural network based framework for sketch generation that does not require ground-truth data for training and produces various styles of sketches. The method combines simple analytic loss functions that correspond to characteristics of the sketch. The network is trained on and evaluated for human face images. Several stylized variations of sketches are obtained by varying the parameters of the loss functions. The paper also discusses the implicit abstraction afforded by the deep convolutional network approach which results in high quality sketch output.


C02: Curb Detection for a Pedestrian Assistance System using End-to-End Learning

Qureshi,H.S., Wizcorek,R.

Our goal is to develop an assistance system for supporting road crossing among older pedestrians. In order to accomplish this, we propose detecting the curb stone from the pedestrians" point of view. To tackle this problems, we chose to fuse two sensors, a Camera and a Leddar, and use an algorithm that applies an end-to-end learning approach. The convolutional neural network was chosen to process the images acquired from the mono camera by filming the curb and its surroundings. The artificial neural network was selected to process the point cloud data of the Leddar acquired in the form of arrays from the 16 channels of the Leddar. A prototype was developed for data collection and testing purposes. It consists of a structure carrying both sensors mounted on a walker. The data from both sensors were collected with multiple factors taken into consideration, such as, weather, light conditions and, approaching angles. For the training of algorithms, an end-to-end learning approach was selected where we labelled the complete image or array rather than labelling the individual pixels or points in the data. The networks were trained and, the features from the parallel networks were concatenated and given as the input to the fully connected layers to train the complete network. The experimental results show an accuracy of more than 99\% and robustness of the end-to-end learning approach. Both sensors are relatively inexpensive and are in fusion together, they are able to efficiently accomplish the task of detecting the curb stone from the pedestrians" point of view.


C11: SyCaT-Vis: Visualization-Based Support of Analyzing System Behavior based on System Call Traces

Hausdorf,A., Hinzmann,N., Zeckzer,D.

Detecting anomalies in the behavior of a computer system is crucial for determining its security. One way of detecting these anomalies is based on the assessment of the amount and sequence of system calls issued by processes. While the number of processes on a computer can become very large, the number of system calls issued during the lifespan of such a process and its subprocesses can be humongous. In order to decide whether these anomalies are due to the intended system usage or if they are caused by malicious actions, this humongous amount of data needs being analyzed. Thus, a careful analysis of the system calls’ types, their amount, and their temporal sequence requires sophisticated support. Visualization is frequently used for this type of tasks. Starting with a carefully aggregation of the data presented in an overview representation, the quest for information is supported by carefully crafted interactions. These allow filtering the tremendous amount of data, thus removing the standard behavior data and leaving the potentially suspicious one. The latter can then be investigated on increasingly finer levels. Supporting this goal-oriented analysis, we propose novel interactive visualizations implemented in the tool SyCaT-Vis. SyCaT-Vis fosters obtaining important insights into the behavior of computer systems, the processes executed, and the system call sequences issued.


C13: Evaluation of 4D Light Field Compression Methods

Barina,D., Chlubna,T., Solony,M., Dlabaja,D., Zemcik,P.

Light field data records the amount of light at multiple points in space, captured by either an array of cameras or by a light-field camera that uses microlenses. Since the storage and transmission requirements for such data are tremendous, compression techniques for light fields are gaining momentum in recent years. Although plenty of efficient compression formats do exist for still and moving images, there is only a little research on the impact of these methods on light field imagery. In this paper, we evaluate the impact of state-of-the-art image and video compression methods on quality of images rendered from light field data. The methods include recent video compression standards, especially AV1 and XVC finalised in 2018. To fully exploit the potential of common image compression methods on four-dimensional light field imagery, we have extended these into three and four dimensions. In the paper, we show that the four-dimensional light field data can be compressed much more than independent still images while maintaining the same visual quality of a perceived picture. We gradually compare the compression performance of all image and video compression methods, and eventually answer the question, "What is the best compression method for light field data?".

C17: Lens-based Focus+Context Visualization Techniques for Interactive Exploration of Web-based Reachability Maps

Listemann, M., Trapp, M., Döllner, J.

Reachability maps are powerful means to help making location-based decisions, such as choosing convenient sites for subsidiaries or planning vacation trips. Existing visualization approaches, however, introduce drawbacks concerning an effective acquisition of relevant information and an efficient data processing. In this paper, we introduce the first approach known so far to apply focus+context techniques to web-based reachability maps. We therefore propose a real-time enabled combination of an isochrone-based and a network-based representation of travel times obtained from multi-modal routing analyses using interactive lenses. We furthermore present a GPU-accelerated image-based method to compute isochrones based on a travel time-attributed network on client-side and thus achieve reduced data transmission efforts while yielding isochrones of higher precision, compared to generalized geometry-based approaches.


C19: Analyzing pixel spread correlation with lenticular lens efficiency on multi user VR displays

Munoz-Arango,Juan S., Reiners,D., Cruz-Neira, C.

One of the all time issues with Virtual Reality systems regardless if they are head-mounted or projection based is that they can only provide perspective correctness to one user. Such limitation affects collaborative work which is nowadays the standard anywhere. One of the approaches for generating different perspective correct images to several users is through optical routing. This approach relies on bending the light to generate perspective correct images to the engaged users. Lenticular lenses can bend the light to be able to generate perspective correct images for several users depending on their positions. On this paper we present an analysis that lets us understand the pixel spread correlation with lenticular lens efficiency on multi user VR displays.


C37: Visualizing Time Series Consistency for Feature Selection


Feature selection is an effective technique to reduce dimensionality, for example when the condition of a system, e.g. a sensor-equipped car, is to be understood from multivariate observations. The selection of variables often involves a priori assumptions about underlying phenomena. To avoid the associated uncertainty, we aim at a selection criterion that only relies on the observations. For nominal data, consistency criteria meet this requirement: a variable subset is consistent, if no observations with equal values on the subset have different output values. Such a model-agnostic criterion is also desirable for forecasting. However, consistency has not yet been applied to multivariate time series. In this work, we propose a visual consistency-based technique for analyzing a time series subset"s discriminating ability with respect to temporal observations. An overview visualization conveys the consistency of output progressions that are associated with comparable observations. Interaction concepts and detail visualizations enable analysts to steer the feature selection process towards inconsistencies. We demonstrate the technique"s applicability based on two real-world usage scenarios. The results indicate that the technique allows analysts to assess the combined discriminating ability of time series without any knowledge about underlying phenomena. It is therefore open to any application that involves multivariate time series for forecasting purposes.


C41: Anatomy-Focused Volume Line Integral Convolution for Brain White Matter Visualization

Schult,T., Klose,U., Hauser,T.-K., Ehricke,H.-H.

3D visualization of volumetric line integral convolution (LIC) datasets has been a field of constant research. So far, most approaches have focused on finding suitable transfer functions and defining appropriate clipping strategies in order to solve the problem of occlusion. In medicine, extensions of the LIC algorithm to diffusion weighted magnetic resonance imaging (dwMRI) have been proposed, allowing highly resolved LIC volumes to be generated. These are used for brain white matter visualization by LIC slice images, depicting fiber structures with good contrast. However, 3D visualization of fiber pathways by volume rendering faces the problem of occlusion of anatomic regions of interest by the dense brain white matter pattern. In this paper, we introduce an anatomy-focused LIC algorithm, which allows specific fiber architectures to be visualized by volume rendering. It uses an anatomical atlas, matched to the dwMRI dataset, during the generation of the LIC noise input pattern. Thus, anatomic fiber structures of interest are emphasized, while surrounding fiber tissue is thinned out and its opacity is modulated. Additionally, we present an adaptation of the orientation-dependent transparency rendering algorithm, which recently has been proposed for fiber streamline visualization, to LIC data. The novel methods are evaluated by application to dwMRI datasets from glioma patients, visualizing fiber structures of interest in the vicinity of the lesion.


C43: Detailed 3D Face Reconstruction from a Single RGB Image

Rotger, G., Moreno-Noguer, F., Lumbreras, F., Agudo, A.

This paper introduces a method to obtain a detailed 3D reconstruction of facial skin from a single RGB image. To this end, we propose the exclusive use of an input image without requiring any information about the observed material nor training data to model the wrinkle properties. They are detected and characterized directly from the image via a simple and effective parametric model, determining several features such as location, orientation, width, and height. With these ingredients, we propose to minimize a photometric error to retrieve the final detailed 3D map, which is initialized by current techniques based on deep learning. In contrast with other approaches, we only require estimating a depth parameter per wrinkle, making our approach fast and intuitive. Extensive experimental evaluation is presented in a wide variety of synthetic and real images, including different skin properties and facial expressions. In all cases, our method outperforms the current approaches regarding 3D reconstruction accuracy, providing striking results for both large and fine wrinkles.


C47: Dynamic Radiosity


In this paper we propose novel radiosity implementation which we called . By storing and updating local form factor matrix of closest to the observer patches we have solved 2 main problems of radiosity algorithm: (1) quadratic complexity of algorithm and thus difficulty of application it to large-scale scenes, and (2) possibility of changing geometry on the fly (dynamic geometry).


C53: Corotated meshless implicit dynamics for deformable bodies

Brunet,J.-N., Magnoux,V., Ozell,B., Cotin,S.

This paper proposes a fast, stable and accurate meshless method to simulate geometrically non-linear elastic behaviors. To address the inherent limitations of finite element (FE) models, the discretization of the domain is simplified by removing the need to create polyhedral elements. The volumetric locking effect exhibited by incompressible materials in some linear FE models is also completely avoided. Our approach merely requires that the volume of the object be filled with a cloud of points. To minimize numerical errors, we construct a corotational formulation around the quadrature positions that is well suited for large displacements containing small deformations. The equations of motion are integrated in time following an implicit scheme. The convergence rate and accuracy are validated through both stretching and bending case studies. Finally, results are presented using a set of examples that show how we can easily build a realistic physical model of various deformable bodies with little effort spent on the discretization of the domain.


C61: Coffee Grading with Convolutional Neural Networks using Small Datasets with High Variance

Wallelign,S., Polceanu M., Jemal T., Buche C.,

Convolutional Neural Networks (CNNs) have been established as a powerful class of models for image recognition problems. Despite their success in other areas, CNNs have been applied only for very limited agricultural applications due to the need for large datasets. The aim of this research is to design a robust CNN model that classifies raw coffee beans into their 12 quality grades using small datasets which have high data variability. The dataset contains images of raw coffee beans acquired in two sets using different acquisition technique under varying illuminations which poses a complex challenge to designing a robust model. To design the model, preprocessing techniques were applied to the input in order to reduce task irrelevant features. But adding the preprocessing techniques did not improve the performance of the CNN model for our dataset. We have also used ensemble methods to solve the high variance that exists in networks when working with small datasets. Finally, we were able to design a model that classifies the beans into their quality grades with an accuracy of 89.01% on the test dataset.


C67: Straight Skeleton Computation Optimized for Roof Model Generation


3D building models are important in several fields, such as urban planning and BIM (Building Information Model). However, enormous labor has to be consumed to create these 3D models. In order to automate laborious steps, a GIS and CG integrated system is proposed for automatically generating 3D building models, based on building polygons (building footprints) on digital maps. Digital maps show most building polygons" edges meet at right angles (orthogonal polygon). In either orthogonal or non-orthogonal polygons, we proposed for automatically generating 3D building models with general shaped roofs by straight skeleton computation, which was proposed through two events. However, some roofs are not created by these two events. In this paper, extended methodologies are proposed for adding degeneration events besides the conventional events, and monotone polygon nodes sorting.


C83: Pi-surfaces: products of implicit surfaces towards constructive composition of 3D objects

Raposo,A.N., Gomes,A.J.P.

Implicit functions provide a fundamental basis to model 3D objects, no matter they are rigid or deformable, in
computer graphics and geometric modeling. This paper introduces a new constructive scheme of implicitly-defined
3D objects based on products of implicit functions. This scheme is in contrast with popular approaches like
blobbies, meta balls and soft objects, which rely on the sum of specific implicit functions to fit a 3D object to a set
of spheres.


C89: Collateral effects of the Kalman Filter on the Throughput of a Head-Tracker for Mobile Devices

Roig-Maimo, MF., Mas-Sanso,R.

We have developed an image-based head-tracker interface for mobile devices that uses the information of the front camera to detect and track the user"s nose position and translate its movements into a pointing metaphor to the device. However, as already noted in the literature, the measurement errors of the motion tracking leads to a noticeable jittering of the perceived motion. To counterbalance this unpleasant and unwanted behavior, we have applied a Kalman filter to smooth the obtained positions.
In this paper we focus on the effect that the use of a Kalman filter can have on the throughput of the interface. Throughput is the human performance measure proposed by the ISO 9241-411 for evaluating the efficiency and effectiveness of non-keyboard input devices. The softness and precision improvements that the Kalman filter infers in the tracking of the cursor are subjectively evident. However, its effects on the ISO"s throughput have to be measured objectively to get an estimation of the benefits and drawbacks of applying a Kalman filter to a pointing device.


D02: Exhibition and Evaluation of Two Schemes for Determining Hypersurface Curvature in Volumetric Data

Hauenstein, J., Newman, T.

Advancements in methodologies for determining 3-dimensional manifold (hypersurface) curvature in volumetric data are presented. Such determinations are requisite in certain shape-based visualization and analysis tasks. The methods explored here are convolution-based approaches. In addition to motivating and describing these methods, an evaluation of their (1) accuracy and (2) computational performance is also presented. That evaluation includes benchmarking on both noise-free and noisy volumetric data.


D03: Grayscale and Color Basis Images

Valery Gorbachev, Elena Kaynarova, Anton Makarov, Elena Yakovleva

Orthogonal transformation of digital images can be represented as a decomposition over basis matrices or basis images. Grayscale and color basis images are introduced. For particular case of DWT found basis wavelet images have a block structure similar to frequency bands of the the DWT coefficients. A steganographic scheme for frequency domain watermarking based on this representation is considered. Presented example of detection algorithm illustrates how this representation can be used for frequency embedding techniques.


D05: Performance Evaluation and Comparison of Service-based Image Processing based on Software Rendering

Wegen, O., Trapp, M., Döllner, J., S. Pasewaldt

This paper presents an approach and performance evaluation of performing service-based image processing using software rendering implemented using Mesa3D. Due to recent advances in cloud computing technology (w.r.t. both, hardware and software) as well as increased demands of image processing and analysis techniques, often within an eco-system of devices, it is feasible to research and quantify the impact of service-based approaches in this domain w.r.t. cost-performance relation. For it, we provide a performance comparison for service-based processing using GPU-accelerated and software rendering.


D07: Accounting for Object Weight in Interaction Design for Virtual Reality

Lykke,J.R., Olsen,A.B., Berman,P., Bærentzen,J.A., Frisvad,J.R.

Interaction design for virtual reality (VR) rarely takes the weight of an object - let alone its moment of inertia - into account. This clearly reduces user immersion and could lead to a break-in-presence. In this work, we propose methods for providing a higher fidelity in interactions with virtual objects. Specifically, we present different methods for picking up, handling, swinging, and throwing objects based on their weight, size, and affordances. We conduct user studies in order to gauge the differences in performance as well as sense of presence of the proposed techniques compared to conventional interaction techniques. While these methods all rely on the use of unmodified VR controllers, we also investigate the difference between using controllers to simulate a baseball bat and swinging a real baseball bat. Interestingly, we find that realism of the motions during interaction is not necessarily an important concern for all users. Our modified interaction techniques, however, have the ability to push user performance towards the slower motions that we observe when a real bat is used instead of a VR controller on its own.


D13: 3D Annotations for Geospatial Decision Support Systems

Zechmeister, S., Cornel, D., Waser, J.

In virtual 3D environments, it is easy to lose orientation while navigating or changing the view with zooming and panning operations. In the real world, annotated maps are an established tool to orient oneself in large and unknown environments. The use of annotations and landmarks in traditional maps can also be transferred to virtual environments. But occlusions by three-dimensional structures have to be taken into account as well as performance considerations for an interactive real-time application. Furthermore, annotations should be discreetly integrated into the existing 3D environment and not distract the viewer"s attention from more important features. In this paper, we present an implementation of automatic annotations based on open data to improve the spatial orientation in the highly interactive and dynamic decision support system Visdom. We distinguish between line and area labels for object-specific labeling, which facilitates a direct association of the labels with their corresponding objects or regions. The final algorithm provides clearly visible, easily readable and dynamically adapting annotations with continuous levels of detail integrated into an interactive real-time application.


D17: Identification of stereo rig alignment error based on vertical disparity map


Poor quality of 3D video content can lead to headache, blurry vision and overall exhaustive experience for the viewer. To ensure quality and comfortable 3D experience for the end consumer, common production errors must be detected and corrected. Vertical disparity is one of these distortions and is caused by improper stereo-camera setup. This paper aims at identifying the possible rotational and placement errors that cause the vertical disparity. An estimation of these errors is necessary to produce good quality 3D content. According to our knowledge, there exists no method to identify rig alignment errors without the knowledge of camera setup parameters and this work is the first step in that direction. Feature point detection has proven to be an interesting approach to find vertical disparity present in the given stereo image pair. In this work feature extraction techniques such as SIFT, SURF and Harris features are efficiently used to compute reliable and robust vertical disparity patterns. This paper classifies vertical disparity patterns according to rig errors. If the vertical disparity values are too small or ambiguous to be identified by pattern analysis, this paper uses graphical analysis that highlights the relationship between the total vertical disparity and the contribution of each possible error to the total. Experimental results show that the proposed approach identifies the reason behind the presence of vertical disparity of some stereo image pairs.


D43: Compressed Exposure Sequences for HDR Imaging

Sekmen, S., Akyüz, A. O.

High dynamic range (HDR) imaging techniques allow photographers to capture the luminance distribution in the real-world as it is, freeing them from the limitations of capture and display devices. One common approach for creating HDR images is the multiple exposures technique (MET). This technique is preferred by many photographers as multiple exposures can be captured with off-the-shelf digital cameras and later combined into an HDR image. In this study, we propose a storage scheme in order to simplify the maintenance and usability of such sequences.
In our scheme, multiple exposures are stored inside a single JPEG file with the main image representing a user-selected reference exposure. Other exposures are not directly stored, but rather their differences with each other and the reference is stored in a compressed manner in the metadata section of the same file. This allows a significant reduction in file size without impacting quality. If necessary the original exposures can be reconstructed from this single JPEG file, which in turn can be used in a standard HDR workflow.


D47: The City Metaphor in Software Visualization

Clinton L. Jeffery

A city metaphor has become a popular method of visualizing properties of program code. This paper provides an overview of research projects that employ this metaphor for a wide range of software engineering tasks. Thusfar projects employing the city metaphor have primarily focused on visualizing static and semi-static properties of software repositories,such as understanding how a program’s source code structure is changing over time, and who is changing what. This paper compares these existing code cities and suggests likely avenues of future research.



D59: Alignment of Point Clouds for Comparison of Infrastructures in Civil Engineering on Quality Control in Metrology

H.Khaksari Haddad, D. Laurendeau

For 3D point cloud registration, Go-ICP cite{yang2016go} has been shown to obtain the global optimal solution for a pair composed of a model point cloud and a data point cloud. Go-ICP mostly has been investigated only on standard sets of point clouds. In this paper, we demonstrate the remarkable efficacy of Go-ICP for the alignment of very complex large-scale point clouds to their corresponding deformed CAD models. In particular, given two distinct sets of point clouds taken from the exterior and the interior of a building, experiments demonstrate that Go-ICP is able to successfully align both of these sets to the point cloud of the CAD model of the whole building (both exterior and interior information included). With the experimentation presented in this paper, we demonstrate that Go-ICP can achieve excellent alignment results and that this approach can be deployed in applications aiming at comparing CAD models of a building ("as designed" model) to the point cloud of the actual building ("as-built" model). Experiments also demonstrate the efficacy of Go-ICP to align a deformed copy of a man-made object to the original object in quality control applications.


D61: Triplet Loss with Channel Attention for Person Re-identification

Organisciak,D., Riachy,C., Aslam,N., Shum, H.P.H.

The triplet loss function has seen extensive use within person re-identification. Most works focus on either improving the mining algorithm or adding new terms to the loss function itself. Our work instead concentrates on two other core components of the triplet loss that have been under-researched. First, we improve the standard Euclidean distance with dynamic weights, which are selected based on the standard deviation of features across the batch. Second, we exploit channel attention via a squeeze and excitation unit in the backbone model to emphasise important features throughout all layers of the model. This ensures that the output feature vector is a better representation of the image, and is also more suitable to use within our dynamically weighted Euclidean distance function. We demonstrate that our alterations provide significant performance improvement across popular re-identification data sets, including almost 10% mAP improvement on the CUHK03 data set. The proposed model attains results competitive with many state-of-the-art person re-identification models.


D67: 3D Gaussian Descriptor for Video-based Person Re-Identification

Riachy,C., Al-Maadeed,N., Organisciak,D., Khelifi,F., Bouridane,A.

Despite being often considered less challenging than image-based person re-identification (re-id), video-based person re-id is still appealing as it mimics a more realistic scenario owing to the availability of pedestrian sequences from surveillance cameras. In order to exploit the temporal information provided, a number of feature extraction methods have been proposed. Although the features could be equally learned at a significantly higher computational cost, the scarce nature of labelled re-id datasets encourages the development of robust hand-crafted feature representations as an efficient alternative, especially when novel distance metrics or multi-shot ranking algorithms are to be validated. This paper presents a novel hand-crafted feature representation for video-based person re-id based on a 3-dimensional hierarchical Gaussian descriptor. Compared to similar approaches, the proposed descriptor (i) does not require any walking cycle extraction, hence avoiding the complexity of this task, (ii) can be easily fed into off-shelf learned distance metrics, (iii) and consistently achieves superior performance regardless of the matching method adopted. The performance of the proposed method was validated on PRID2011 and iLIDS-VID datasets outperforming similar methods on both benchmarks.





A31: Tools for development of interactive web-based maps: application in healthcare

Karolyi,M., Krejcí,J., Štavnický,J., Vyškovský,R., Komenda,M.

Interactive visualisations on the Internet have become commonplace in recent years. Based on such publicly available visualisations, users can obtain infor-mation from various domains quickly and easily. A location-specific method of data presentation can be much more effective using map visualisation than using traditional methods of data visualisation, such as tables or graphs. This paper pre-sents one of the possible ways of creating map visualisations in a modern web environment. In particular, we introduce the technologies used in our case togeth-er with their detailed configuration. This description can then serve as a guide for the customisation of the server environment and application settings so that it is easy to create the described type of visualisation outputs. Together with this man-ual, specific cases are presented on the example of an application which was de-veloped to display the location of medical equipment in the Czech Republic based on data collected from healthcare providers.


A79: Porting A Visual Inertial SLAM Algorithm To Android Devices

Möller,J., Meyer,B., Eisemann,M.

Simultaneous Localization and Mapping aims to identify the current position of an agent and to map his surroundings at the same time. Visual inertial SLAM algorithms use input from visual and motion sensors for this task. Since modern smartphones are equipped with both needed sensors, using VI-SLAM applications becomes feasible, with Augmented Reality being one of the most promising application areas. Android, having the largest market share of all mobile operating systems, is of special interest as the target platform. For iOS there already exists a high-quality open source implementation for VI-SLAM: The framework VINS-Mobile. 
In this work we discuss what steps are necessary for porting it to the Android operating system. We provide a practical guide to the main challenge: The correct calibration of device specific parameters for any Android smartphone. We present our results using the Samsung Galaxy S7 and show further improvement possibilities.


D71: Techniques for GPU-based Color Palette Mapping

Trapp, M., Pasewaldt, S., Döllner, J.

This paper presents a GPU-based approach to color quantization by mapping of arbitrary color palettes to input images using Look-Up Tables (LUTs). For it, different types of LUTs, their GPU-based generation, representation, and respective mapping implementations are described and their run-time performance is evaluated and compared.


B07: Open Data Visualization in Danish Schools: A case Study

Saddiqa, M., Rasmussen, L., Larsen, B., Magnussen, R., Pedersen, J.M.

Increasingly public bodies and organizations are publishing Open Data for citizens to improve their quality of life and solving public problems. But having Open Data available is not enough. Public engagement is also important for successful Open Data initiatives. There is an increasing demand for strategies to actively involve the public exploiting Open Data, where not only the citizens but also school children and young people are able to explore, understand and extract useful information from the data, grasp the meaning of the information, and to visually represent findings. In this research paper, we investigate how we can equip our future generation with the skills to understand and make use of Open Data. We present the results of a survey among Danish school teachers and pupils. The survey focuses on how we can introduce interactive visualization using Open Data, and which competencies are needed for pupils and teachers to understand, use and present the data. We briefly review Copenhagen city’s Open Data and existing open source software suitable for visualization, to study which open source software pupils can easily adapt to visualize Open Data and which data-sets teachers can relate to their teaching themes. Our study shows that introducing Open Data visualization can bring fruitful outcomes in everyday teaching and that to actively use Open Data in schools, teachers and pupils need to boost their digital skills.


B29: Prototyping a game engine architecture as a multi-agent system

Marín,C., Chover,M., Sotoca,J.M.

The game engines are one of the essential and daily used applications on the game development field. These applications are designed to assist in the creation of this type of contents. Nevertheless, their usage and development are very complex. Moreover, there are not many research papers about the game engine architecture definition and specification. In this sense, this work presents a methodology to specify a game engine defined as a multi-agent system. In such a way, from a multi-agent approximation, a game engine architecture can be prototyped in a fast way. Also, this implementation can be exported to a general programming language for maximum performance, facilitating the definition and comprehension of the mechanisms of the game engine.



Dubrovskaya, M

Here is proposed a solution for actual problem of heart allocation on tomography data. One of the most popular tomography types used for heart is CT, thus current technology, proposed here, is initially developed for CT data. The main part of the algorithm consists of four stages: preliminary source volume limitation, forming hypotheses of lungs location, determination of approximate surrounding heart cube and searching for a point within the cube which is for sure inside the heart. The algorithm also includes a preliminary stage: coordinate system allocation. This stage is essential when position of a scanned patient is unknown relative to a tomograph. During the stage we find, where left and right, top and bottom, front and back sides of human body are. The stage consists itself also of four parts: definition of Ox and Oy axes and determination of Oy, Oz and Ox directions, by analyzing anatomy of a patient, with regard to standard human anatomy. Method is fast, doesn`t require any human guidance and preliminary adjustment. The whole method is convenient as automatic preprocessing for unlimited scope of problems, e.g. segmentation, surface or solid model reconstruction, etc. Proposed method provides flexible way for further local heart positioning refinement and works even for static, not timed, tomography data.



B83: Real Time Pedestrian and Object Detection and Tracking-based Deep Learning. Application to Drone Visual Tracking

R. Khemmar, M. Gouveai, B. Decoux, JY. Ertaud

This work aims to show the new approaches in embedded vision dedicated to object detection and tracking for drone visual control. Object/Pedestrian detection has been carried out through two methods: 1. Classical image processing approach through improved Histogram Oriented Gradient (HOG) and Deformable Part Model (DPM) based detection and pattern recognition methods. In this step, we present our improved HOG/DPM approach allowing the detection of a target object in real time. The developed approach allows us not only to detect the object (pedestrian) but also to estimates the distance between the target and the drone. 2. Object/Pedestrian detection-based Deep Learning approach. The target position estimation has been carried out within image processing. After this, the system sends instruction to the drone engine in order to correct its position. For this visual servoing, we have developed two kinds of PID controllers. The platform has been validated under different scenarios by comparing measured data to ground truth data given by the GPS of the drone. Several tests which were carried out at ESIGELEC car park and Rouen city center validate the developed platform.


B89: Title Segmentation in Arabic Document Pages

Bouressace Hassina

Recent studies on text line segmentation have not focused on title segmentation in complex structure documents, which may represent the upper rows in each article of a document page. Many methods cannot correctly distinguish between the titles and the text, especially when it contains more than one title. In this paper, we discuss this problem and then present a straightforward and robust title segmentation approach. The proposed method was tested on PATD (Printed Arabic Text Database ) images and we achieved good results.


C59: CNN Approaches for Dorsal Hand Vein Based Identification

Lefkovits, S., Lefkovits, L., Szilágyi, L.

In this paper we present a dorsal hand vein recognition method based on convolutional neural networks (CNN). We implemented and compared two novel CNNs trained from end-to-end to the most important state-of-the-art deep learning architectures (AlexNet, VGG, ResNet and SqueezeNet). We applied the transfer learning and fine-tuning techniques for the purpose of dorsal hand vein-based identification. The experiments carried out studied the accuracy and training behaviour of these network architectures. The system was trained and evaluated on the best-known database in this field, the NCUT, which contains low resolution, low contrast images. Therefore, different pre-processing steps were required, leading us to investigate the influence of a series of image quality enhancement methods such as Gaussian smoothing, inhomogeneity correction, contrast limited adaptive histogram equalization, ordinal image encoding, and coarse vein segmentation based on geometrical considerations. The results show high recognition accuracy for almost every such CNN-based setup.


D23: Improving facial attraction in videos

Silva,M., Martino,J.

The face plays an important role both socially and culturally and has been extensively studied especially in investigations on perception. It is accepted that an attractive face tends to draw and keep the attention of the observer for a longer time. Drawing and keeping the attention is an important issue that can be beneficial in a variety of applications, including advertising, journalism, and education. In this article, we present a fully automated process to improve the attractiveness of faces in images and video. Our approach automatically identifies points of interest on the face whimeasures the distances between them while adding the use of classifiers to identify the pattern of points of interest most adequate to improve the attractiveness.
The modified points of interest are projected in real-time in three-dimensional face mesh to support the consistent transformation of the face in a video sequence. In addition to the geometric transformation, texture is also automatically smoothed through a smoothing mask and weighted sum of textures. The process as a whole enables the improving of attractiveness not only in images but also in videos in real time.


D29: Obstacle detection algorithm by means of images with a ZED camera using ROS software in a drone

Pérez Gutierrez, M., Córdova-Cruzatty, A.

During flight conditions, Unmanned Air Vehicles (UAVs) can be in situations of high risk such as encountering unforeseen obstacles, thus resulting in physical damage of the device and its equipment. For this reason, an intelligent pilot system that allows the detection and evasion of obstacles is necessary for UAVs, protecting the integrity of the system without compromising the mission or the established flight plan. In order to implement this intelligent pilot system, a fundamental element was used, a ZED camera that acts in a similar way to people"s eyes and is integrated with ROS software for depth detection through an obstacle algorithm.
The interest of this research is focused on the obstacle detection algorithm using a stereoscopic camera capable of creating 3D images and software that contributes to develop applications for robots and research platforms. Once the detection algorithm was implemented, it was tested in two specific stages, the first stage was established to find the optimum detection distance through the establishment of distance ranges and a second stage included controlled and uncontrolled environments in order to carry out the algorithm validation.


D31: Immersive Analytics Sensemaking on Different Platforms

Blum,S., Cetin,G., Stuerzlinger,W.

In this work we investigated sensemaking activities on different immersive platforms. We observed user behaviours during a classification task on a very large wall-display system (experiment I) and in a modern Virtual Reality headset (experiment II). In experiment II, we also evaluated a condition with a VR headset with an extended field of view, through a sparse peripheral display. We evaluated the results across the two studies by analyzing quantitative and qualitative data, such as task completion time, number of classifications, followed strategies, and shape of clusters. The results showed differences in user behaviours between the different immersive platforms, i.e., the very large display wall and the VR headsets, and also point to several opportunities for future research.







B73: The Role of Polarization in The Multiple Reflections Modeling

Basov,A.Y., Budak,V.P., Grimailo,A.V.

In this article, a new mathematical model for multiple reflections calculation based on the Stokes vector and Mueller matrices is proposed. The global illumination equation and local estimations method were generalized on the polarization case. Results of the multiple reflections calculation with using of local estimations method show the difference of more than 30% between standard calculation and polarization-accounting one. The way of describing the surface reflection with polarization account is proposed.