Search

Recommendation systems are present in many daily activities. They are trying to predict user preferences. Due to the growth of social networks, there is a vast amount of data that is constantly updated which makes recommendation systems more personalized and efficient. This study aims to apply natural language processing (NLP) and deep learning techniques to obtain a recommendation. NLP is used to analyze the text (i.e. hashtags) from social networks to determine similarity between different points of interest (POI). A pre-trained convolutional neural network (CNN) is used to classify a set of images obtained from social networks to determine which POI is visited by which personality type. The personality type is determined using the Five-Factor (that is, Big Five) model. The Big Five traits are firstly converted into ten personality class labels (High Openness, Low Openness, High Conscientiousness, Low Conscientiousness, High Extraversion, Low Extraversion, High Agreeableness, Low Agreeableness, High Neuroticism, Low Neuroticism) for the classification network. We manually labeled more than 2, 000 images and used a pre- trained CNN in a transfer learning manner to automatically extract features from images and classify them. We demonstrated that personality traits can be extracted from posted images with an accuracy of 75%. Also, we showed that those traits can be aggregated for a given set of pictures, such that a representation of a destination can be determined.

As the title suggests, this presentation is all about the most popular tools used in various phases of software development, as well as education. Not to reveal too much, this will be a brief journey through the world of tools that make life easier to developers, revealing several modern tools and their usage in every day's development. Stay tuned?

This community keynote brought to you by HUJAK, will present the past, the present and, most important, the future of Java as one of the best programming languages and the most widespread software development platform in the world today, but, even more important, as an enormous ecosystem of code, libraries, tools, services, and ? the people behind it all. We will take on a journey through the most interesting features of the Java programming language in the last years, as well as the most prospective new projects and features that already are or will be a part of the future of Java. To name a few, we will discuss many contemporary projects such as Valhalla, Panama, Graal, Loom, and Metropolis, as well as new features related to Switch Expressions, Text Blocks, Pattern Matching, NullPointerExceptions, Records, Foreign Function and Memory API, Vector API, and Sealed Classes. We will also address the current state of software development related to Java and similar technologies, presenting results from various developer surveys, with additional focus on our part of the world and Croatia. We will also provide recommendations coming from HUJAK as a local Java community and how to respond to current software development challenges, and discuss the future of our software development in the context of remote work, education, and employment.

In this paper, we propose and analyze the novel approach for contactless surgery. The base of contactless surgery is the system for motion tracking. For motion tracking, we use a complex stereo vision depth camera system. Once having the patient data, by using the VR techniques and depth tracking cameras, it is possible to analyze and segment the region of interest in pre-operative preparation. Using a depth tracking camera enables an operation to be simulated, and the outcome is viewed and analyzed before the patient undergoes surgery in the same environment as in the real surgery. This enables medical specialists to optimize their surgical approaches. We think specific VR/AR technologies and approaches should be incorporated, such as analyzing 3D workflow and 3D virtual scene so medical specialists would be able to make the most of the benefits during the surgery. A VR-based application is useful due to avoiding any dangers for the patients. Training applications usually have an elevation part for any step of training. However, the main challenge is to produce a step with adequate fidelity to be simulated like what is performing on the patients. The other ability of training and education simulators is preparing a situation for research surgeons to simply practice what has been suggested by other colleagues and even reading from websites.

Nowadays, there are many new approaches and techniques in telemedicine and surgery with different kinds of innovations and a growing need for contactless control of surgery parameters. Our proposal is aiming to resolve the problem of standard surgical parameters with gesture- controlled surgical interventions. We designed a contactless interface as a plug-in application for the DICOM viewer platform using a hardware sensor device controller that supports hand/finger motion as input, with no hand contact, touching, or voice navigation. Our proposed approach enables surgeons to get complete and aware orientation in the operative field (which currently isn’t the case and where the problem lies), where ‘overlapping’ the real and virtual anatomic models is inevitable. Human mind and understanding of this new surgery work by creating entirely new models of human behavior and understanding spatial relationships, along with devising assessment that will provide an insight into our human nature. That’s why the essential part of the solution is to build the adequate hand and gesture module for motion which we will describe in this paper.

Full-Text HTML Abstract Full-Text PDF Full-Text XML How to Cite Research Article What is the Future of Minimally Invasive Sinus Surgery: Computer-Assisted Navigation, 3D-Surgical Planner, Augmented Reality in the Operating Room with ‘in the Air’ Surgeon’s Commands as “Biomechanics” of the New Era in Personalized Contactless Hand- Gesture Non-Invasive SurgeonComputer Interaction? Klapan Ivica1, 2, 3, 4*, Duspara Alen6, Majhen Zlatko5, 7, Benić Igor8, Kostelac Milan8, Kubat Goranka9, Berlengi Nedjeljka10, Zemba Mladen10, Žagar Martin11 Author Affiliations Received: July 11, 2019 | Published: July 23, 2019 Corresponding author: Ivica Klapan, Klapan Medical Group Polyclinic, Ilica 191A, HR-10000 Zagreb, Croatia, EU DOI: 10.26717/BJSTR.2019.19.003377 Abstract Purpose: We were focused on the development of personal-3D-navigation system and application of augmented reality in the operating room per viam personalized contactless hand-gesture non-invasive surgeon-computer interaction, with higher intraoperative safety, reduction of operating time, as well as the length of patient postoperative recovery. Methods: Simultaneous use of video image, 3D anatomic fields and navigation in space, with the application of our original special plug-in application for OsiriX platform, enabling users to use LM-sensor as an interface for camera positioning in 3DVR and VE views, and integrating speech recognition as a VC solution, in an original way. Results: Management of image 2D-3D-video-medical documentation, as well as the control of marker- based virtual reality simulation in real time during real operation with per viam our personalized contactless “in the air” surgeon’s commands. Conclusion: The use of modern technologies in head and neck surgery in the last 30 years (e.g., FESS, NESS, and robotic surgery) has enabled surgeons to demonstrate spatial anatomic elements in the operating field, which was quite inconceivable before. This approach has not yet been used in rhinosinusology or otorhinolaryngology, and to our knowledge, not even in general surgery. The question that we have to ask ourselves now is what prerequisites, realizable in the future, should be realized in our “on the fly” gesturecontrolled and incisionless virtual surgical interventions for the eventual utilization to meet the most demanding requirements in the OR?

Objectives: This study presents the use of our original contactless interface as a plug-in application for OsiriX-DICOM-viewer platform using a hardware sensor devicecontroller that supports hand/finger motions as input, with no hand contact, touching, or voice navigation. It would be possible to modify standard surgical parameters in the fly gesture-controlled incisionless surgical interventions. Methods: The accuracy of computer-generated models was analyzed according to T. Galeta/2017. Our original special plug-in- application provided different types of gestures for three-dimensional- virtual reality navigation. Our hardware sensor device controlling the system without touching any other device served as an interface for camera positioning in three- dimensional virtual endoscopy-views. The impression of panoramic three-dimensional volume rendering-viewing was given by pivoting the camera around a focus fixed on the object. Results: This novel technique enables surgeons to get complete and aware orientation in the operative field, where ‘overlapping’ of the real and virtual anatomic models is inevitable. Our human mind and understanding of this new surgery work by creating completely new models of human behavior and understanding spatial relationships, along with devising assessment that will provide an insight into our human nature. Any model and/or virtual model of the surgical field is defined as it actually exists in its natural surroundings. Conclusion: We offer an alternative to closed software systems for visual tracking, with an initiative for developing the software framework that will interface with depth cameras, with a set of standardized methods for medical applications such as hand gestures and tracking, face recognition, navigation, etc. This software should be an open source, operation system agnostic, approved for medical use and independent of hardware. Comparison with previous doctrine in human medicine clearly indicates that both preoperative/intraoperative manipulation with three-dimensional-volume rendering slices of the human anatomy per viam touchless surgical navigation system with simulation of virtual activities has become reality in the operation room.

V diplomski nalogi sem obravnaval ekonomično debelino toplotne izolacije na zunanji steni ter pod temeljno ploščo. Poleg tega pa sem primerjal ceno gradnje klasične in montažne hiše. V prvem delu sem predstavil lastnosti klasično zidane in montažne hiše ter opredelil teoretične osnove energetske učinkovitosti. V drugem delu sem opisal izbrani objekt ter zanj izdelal popis materialov za primera klasično zidane hiše in montažne hiše, pri čemer sem za primerjavo vzel montažno hišo Marles s sestavo zunanje stene Mega N14. Na podlagi izbranega objekta sem izračunal ekonomično debelino toplotne izolacije zunanjih sten v obdobju 60 let, kjer se fasada obnovi dvakrat in ekonomično debelino toplotne izolacije pod temeljno ploščo pri predpostavljeni življenjski dobi 100 let za primer klasično zidane hiše. Upošteval sem tudi možnost, da se toplotna izolacija pod temeljno ploščo ne izvede. Za izračun sem uporabil dinamično metodo iz ekonomije, imenovano neto sedanja vrednost. V nadaljevanju sem izdelal popis del in materialov za izvedbo izbranega objekta ter tako primerjal ceno izvedbe klasične in ceno izvedbe montažne hiše. Pri tem sem zaradi primerljivosti določene postavke klasično zidane hiše poenotil s ponudbo montažne hiše Marles Mega N14, tako da sta oba objekta enaka v smislu energetske varčnosti ter v smislu enake opreme. V diplomskem delu sem prikazal, da se vgrajevanje večje količine toplotne izolacije obrestuje tako na zunanjih stenah kot pod temeljno ploščo. Rezultati cenovne primerjave klasično zidane hiše in montažne hiše so pokazali, da je izvedba zidanega objekta še vedno cenejša za približno 20 %. In the thesis, I discussed the economic thermal insulation thickness on the outer wall and under the base plate. I also compared the prices of constructing a classically built and prefabricated house. In the first part, I presented characteristics of the classically built and the prefabricated house and defined the theoretical bases of energy efficiency. In the second part, I described the chosen object and prepared an inventory of the material for the cases of a classically built house and a prefabricated one, where I compared the prefabricated house by Marles with the N14 composition of the outer wall. Based on the chosen object I calculated the economic thermal insulation thickness of the outer walls for a period of 60 years, where the façade is being renovated twice, and the economic thermal insulation thickness under the base plate with the assumed lifetime of 100 years for the case of a classically built house. I also considered the possibility of no thermal insulation under the base plate. For the calculation, I used the dynamic method, used in economics, called net present value. In continuation I prepared an inventory of the work and material for the construction of the chosen object and so compared the prices for the construction of a classically build and a prefabricated house. For the purpose of comparability, I unified some elements of the classically built house with the offer for the prefabricated house Marles Mega N14, so that both objects are the same in terms of energy efficiency and equipment. In the thesis, I showed that the installation of large amounts of thermal insulation pays off at the outer walls as well as under the base plate. The results of the price comparison for the classically built house and the prefabricated one showed that the construction of a classically built object is still cheaper for about 20%.

Lossless compression of visual data is required in applications such as medical imaging, image archival, remote and satellite imaging etc. We present a new adaptive predictive image coding method based on the blending of multiple static predictors on a dynamically classified causal context of neighboring pixels. The idea of predictor blends is further expanded through the determination of blending context that changes its shape on a pixel--by--pixel basis using simple classification technique, thus allowing the modeling more complex image structures such as nontrivially oriented edges, the periodicity and coarseness of textures. Typical natural images are characterized as being composed of image regions with different local properties. The predictor estimates those properties around the currently unknown pixel and adjusts itself so that the presence of detected properties affect the way the final prediction is made. Predictive part is followed by a heuristic contextual model and statistical encoder.

U ovom radu predložen je novi programski okvir za kompresiju četvero-dimenzionalnih (4D) medicinskih podataka. Arhitektura ovog programskog okvira temelji se na različitim procedurama i algoritmima koji detektiraju vremenske i prostorne zalihosti u ulaznim 4D medicinskim podacima. Pokret kroz vrijeme analizira se pomoću vektora pomaka koji predstavljaju ulazne parametre za neuronske mreže koje se koriste za procjenu pokreta. Kombinacijom segmentacije, pronalaženja odgovarajućih blokova i predikcijom vektora pomaka, zajedno s ekspertnim znanjem moguće je optimirati performanse sustava. Frekvencijska svojstva se analiziraju proširenjem wavelet transformacije na tri dimenzije. Za mirne volumetrijske objekte, moguće je konstruirati različite wavelet pakete s različitim filtrima koji omogućavaju širok raspon analiza frekvencijskih zalihosti. Kombinacijom uklanjanja vremenskih i prostornih zalihosti moguće je postići vrlo visoke omjere kompresije., This work presents a novel framework for four-dimensional (4D) medical data compression architecture. This framework is based on different procedures and algorithms that detect time and spatial (frequency) redundancy in recorded 4D medical data. Motion in time is analyzed through the motion fields that produce input parameters for the neural network used for motion estimation. Combination of segmentation, block matching and motion field prediction along with expert knowledge are incorporated to achieve better performance. Frequency analysis is done through an extension of one dimensional wavelet transformation to three dimensions. For still volume objects different wavelet packets with different filter banks can be constructed, providing a wide range of frequency analysis. With combination of removing temporal and spatial redundancies, very high compression ratio is achieved.

GML (eng. Geography Markup Language) je oblikovan da podrži interoperabilnost i to omogućava kroz uporabu osnovnih geometrijskih oznaka - tagova (svi sustavi koji podržavaju GML koriste iste geometrijske tagove), zajedničkim podatkovnim modelom (svojstvima) i mehanizmom za kreiranje i dijeljenje aplikacijskih shema. Kratki uvod u GML se nalazi u prvom poglavlju. Međutim, GML datoteke ne sadržavaju informacije o stilu i načinu prikazivanja. To se može riješiti na više načina, od kojih je za ovaj diplomski rad izabrano rješenje vizualizacije pomoću SVG-a (eng. Scalable Vector Graphics). Opis SVG standarda se nalazi u trećem poglavlju. Za prikaz je potrebno format GML transformirati u format SVG. Transformacija se obavlja definiranjem pripadne liste pravila, odnosno pripadnog XSLT dokumenta. Postupci stvaranja XSLT dokumenta su opisani u četvrtom poglavlju. U petom poglavlju je dan primjer prikaza razdjelne mreže u GML formatu.