Application of PyTorch3D and NERF computer vision tools for building a point cloud of a three-dimensional model and determining the camera position of still images in space

V.V. Konkov, A.B. Zamchalov

Accepted: 2024-10-03

Abstract

Recently, computer graphics plays a key role in solving computer vision problems. The problem of converting 2D images into 3D models continues to be urgent, as it requires precise determination of camera position and construction of accurate 3D models of objects. Traditional methods are often limited in application and do not offer a comprehensive solution. This study examines the use of PyTorch3D and NERF libraries to determine the camera position in 3D space and create a 3D model of an object from a single 2D image. As a method of data preparation, a hardware and software system was used, including a stepper motor control device that provides manual and sequential positioning of the camera and its return to the initial position, a shooting control system to generate a comprehensive set of photos at each camera position, and a mechanism for sending data to a remote computer for further processing. The PyTorch3D library was selected during the study to explore the possibilities of converting 2D images into 3D models or determining the position of an object in the photos. The processing process included several steps: building a point cloud to generate a 3D volumetric model of the object, determining the camera position in 3D space from a single 2D image using inverse problem algorithms, and constructing a 3D object using differentiable rendering, creating 3D voxels and 3D meshes. The results of this study showed successful determination of camera position in 3D space and construction of a 3D object model from a single 2D image, demonstrating the advantages of using the PyTorch3D library over other existing models. These findings can be applied in the development of software and hardware systems for creating 3D images from 2D photographs. The study confirmed the relevance and effectiveness of using PyTorch3D library to solve the problems of converting 2D images into 3D models. Further work will be aimed at expanding the functionality of the system and its use in various areas of computer vision.

Visualization of methods of machine learning. GUI programming

Accepted: 2024-09-24

Abstract

The technologies of artificial intelligence and machine learning have made a fundamental leap in their capabilities in the last five years. The growth of processing power and the emergence of more and more effective methods of machine learning allows AI to not just solve the most typical tasks associated with the field, such as statistical analysis and optimization of mathematical processes, but also to find new applications in related fields of research, as well as practical applications, including those on the free market, available to the mass consumer. Image generation, audio, animation, self-learning models of control of robotic platforms and virtual mechanical models – these and many more novel applications of the recent years have led to a media-boom around AI and a growing interest from developers and authors from various fields and industries.
That being said, the methods for developing, research, testing, and integration of AI have largely remained unchanged and still require the knowledge of programming languages, machine learning libraries, as well as a deep understanding and experience specifically in the narrow field of AI. This barrier of specialization not only demands inclusion of machine learning specialists in the development process of otherwise trivial computer applications, typical for the field of AI, but also prevents small teams and independent developers from using the latest advances in these technologies without significant monetary and time investments into studying the subject.
We offer a novel solution to this issue in the form of a prototype graphical interface that allows the user without technical education and without the need for knowledge of programming languages to develop and tune various architectures of neural nets and other machine learning methods, methods of unsupervised machine learning, and to test these methods on a wide range of experimental tasks – from mathematical equations to controlling virtual mechanical models in a simulated physical environment. In this article, we give a brief description of its structure and organisation, its fundamental principles of operation, and the capabilities of this GUI.

Use of Hadamard matrices in single-pixel imaging

Denis V. Sych

Accepted: 2024-08-13

Abstract

Single-pixel imaging is a method of computational imaging that allows to obtain images of objects using a photodetector that does not have spatial resolution. In this method, the object is illuminated by light having a special spatio-temporal structure, — light patterns, and a single-pixel photodetector measures the total amount of light reflected from the object. The possibility of obtaining an image and the image quality are closely related to the properties of the applied patterns and computational algorithms. In this paper, we consider patterns obtained from modified Hadamard matrices and study the features of image calculation using single-pixel imaging. We show the possibility of reducing both the sampling time and the computational resources required to obtain images by modifying the pattern system. The proposed theoretical method can be used in the practical implementation of the single-pixel imaging method in an experiment.

Numerical visualization of vortex wakes behind large particles

А.А. Mochalov, А.Yu. Varaksin

Accepted: 2024-08-11

Abstract

An attempt has been made to visualize the flow formed in the wake of large particles moving in an ascending turbulent air flow in the channel. Numerical modeling was performed using a simplified version of the approach called "two–way coupling" (TWC) in English literature and taking into account the inverse effect of particles on gas characteristics. The particle motion was calculated in an approximate manner, therefore the method used is called "quasi – two–way coupling", TWC(Q). The results of numerical modeling of the characteristics of turbulent trails behind large moving particles based on the Reynolds averaged Navier-Stokes equations (RANS) are presented.

Deep Learning for Effective Visualization and Classification of Recyclable Material Labels

V.O. Kuzevanov, D.V. Tikhomirova

Accepted: 2024-08-06

Abstract

This paper presents an example of a system to improve the process of sorting recyclables by using deep learning techniques to automatically detect, classify and visualize recycling codes on product packages. In this paper, the author discusses various approaches to optical character recognition and object detection in a video stream or image. The author has developed and proposed a combination of neural networks for detection and classification of recycling codes. The proposed neural network system is designed to facilitate efficient recycling processes by automating the identification of recycling symbols, thereby facilitating the sorting and processing of recyclables.

Modeling and visualization of complex shaped surfaces using interpolation curves

E.V. Konopatskiy, A.A. Bezdytniy

Accepted: 2024-08-03

Abstract

This article presents an approach to modeling and visualizing surfaces of complex shapes using interpolation curves with predetermined geometric properties. A modified Bezier curve of -order was used as interpolation curves. Modification of a Bezier arc into an interpolation curve is possible both with and without preserving tangents. When preserving tangents, the Bezier arc retains its properties as a contour arc and acquires the ability to pass through pre-set points. The considered modification is possible in several variations: universal, based on the uniform distribution of the parameter during the modification process, and adaptive, when the parameter values are adapted to the initial data. The use of interpolation curves makes it possible to implement a special case of the moving simplex method, an analogue of which in geometric modeling and computer-aided design systems is the section operation (or lofting). The difference is that a continuous curve is used as a generating surface instead of a piecewise one. This approach has its advantages and disadvantages. For the demonstration, examples of surface models of an onion dome and a vase using various guides are given. The results obtained were compared. The introduction of research results into CAD/CAM will significantly expand their tools in terms of shape formation and visualization of surfaces and bodies that have predetermined geometric requirements.

Visualization metaphors in the tasks of exploratory analysis of heterogeneous data

R.A. Isaev, A.G. Podvesovsky, A.A. Zakharova

Accepted: 2024-08-03

Abstract

The subject of the study is the construction and application of visual models using the concept of visualization metaphors in the context of exploratory analysis of heterogeneous data. This study considers improved variants of the previously proposed visualization metaphors that can be used as a basis for building visual models. A technology for exploratory analysis of heterogeneous data based on the joint use of different visualization metaphors is proposed. The process of visual data exploration at the stage of exploratory analysis using the proposed technology is demonstrated to be iterative and multiscenary, contingent upon the analysis goals. The software tool developed to implement the proposed technology is described, along with its additional functionality to calculate and export quantitative characteristics of the visual model. The software tool is then considered in the context of exploratory analysis of a synthetic data set. The future direction of the proposed approach to the construction of visual models, the technology of exploratory data analysis and the software tool for its support are determined.

Visualizing the Impact of Machine Learning on Cardiovascular Disease Prediction: A Comprehensive Analysis of Research Trends

Jeena Joseph, K Kartheeban

Accepted: 2024-08-01

Abstract

Cardiovascular diseases (CVDs) continue to have a negative impact on global health, which highlights the need for accurate and efficient prediction methods. Machine learning (ML) techniques as tools for forecasting CVD has recently showed potential. This paper presents a comprehensive analysis of research trends in the field, focusing on visualizing the impact of ML in cardiovascular disease prediction. We used data visualization techniques to identify patterns and trends in an extensive database of scholarly publications on this subject that were published in Scopus between 1991 and 2023. The analysis reveals a substantial growth in research output, demonstrating the growing demand for ML-based CVD prediction. It reveals essential stakeholders and potential collaborators while highlighting the institutions and authors who have contributed most to this domain. The study also identifies high-impact journals that have published significant research in this domain, facilitating researchers in selecting appropriate outlets for dissemination. The study helps researchers identify the most critical areas for further research and fosters cooperation among subject-matter experts by offering insightful information about machine learning-based cardiovascular disease prediction development. The data is analyzed using the tools VOSviewer and Biblioshiny.

Exploring the Research Landscape of Business Applications of Robotic Process Automation Through Bibliometric Analysis

Shamini James, S. Karthik, Binu Thomas

Accepted: 2024-08-01

Abstract

The field of business process optimization and automation has seen the emergence of robotic process automation (RPA) as a disruptive technology. This research aims to give a systematic bibliometric analysis of the research ecosystem of robotic process automation in business to identify trends, patterns, and developments in this quickly developing area. Bibliometric methodologies, such as co-authorship analysis, keyword analysis, citation patterns, and publishing trends are performed in this work. Research papers from Scopus scientific databases are incorporated into the analysis through the identification of key writers, organizations, and nations that have made a substantial contribution to the growth of RPA literature. The report also explores the temporal evolution of RPA research, highlighting the development of research areas over time and identifying pockets of active research as well as prospective paradigm shifts. The research reveals key publications that have significantly influenced the course of RPA research by looking at citation networks.
The results of this bibliometric analysis enable scholars, practitioners, and policymakers to develop a more detailed grasp of the RPA research landscape. This study provides a roadmap for future research directions in robotic process automation by identifying research gaps and emerging trends.

Mapping the Knowledge Base on Visual Reality Technology and the Manufacturing Industry

Geofrey Rwezimula, Zhang Guoxing, Wakara Ibrahimu Nyabakora

Accepted: 2024-08-01

Abstract

Virtual reality applications provide users with more than just realistic sight; they may also sense touch, hear, and even interact with virtual objects. With these significant advancements, virtual reality has seen recent growth surges in a number of sectors, including the manufacturing industry. It has to be successful in drawing attention from both academics and industry. It needs to be known how researchers are interested in the technology application. Therefore, examining the body of research on the connection between visual reality and the manufacturing industry is the goal of this research. The bibliometric study was carried out using the Scopus database. Using PRISMA, the sample procedure was finished. VOSviewer was utilized to search through 2,037 publications. This disclosed the expansion of the network, the most active contributing stakeholders, the backdrop of the intellectual framework, the research gap, and the greatest popular topic that needed to be filled. We observed that starting in 1992, papers pertaining to the influence of virtual reality on the manufacturing industry collected from the Scopus database were included. The words “augmented reality,” “virtual reality,” “process simulation,” “industrial internet of things,” “industry 4.0 technologies,” and “3D technologies" have been widely used since 1992. The density map's representation of contemporary themes includes “artificial intelligence” and “human-robot interaction.” The significance of the findings for researchers lies in their relevance to the past, present, and future, along with the identification of knowledge gaps.

Spectral evaluation of the vital state of Quercus robur L. under simulated drought conditions

P.A. Zybinskaya, A.V. Tretyakova, P.A. Krylov

Accepted: 2024-07-24

Abstract

Non-destructive spectral methods of analysis are increasingly being used to study the content of plant metabolites, evaluate morpho-physiological and biochemical indicators, as well as evaluation of the vital state. Visualization of the vital state through spectral profiles can provide a more detailed picture of plant adaptation to stress. To model experimental drought, 5-6 month-old Quercus robur L. seedlings were divided into three groups: control and experimental groups with and without watering (drought), with 15 seedlings in each group. Spectral evaluation of leaf blades was performed using a portable spectroradiometer SpectraPen SP110 Uvis and a plant analyzer Dualex Scientific+ at 0 (control), 168 (one week), and 336 (two weeks) hours. As a result of spectral analysis, spectrograms of radiation absorption of Q. robur leaf blades were obtained, as well as the content of the sum of chlorophylls, flavonols and anthocyanins under watering and drought conditions. The study revealed changes in the spectrograms of absorption of Q. robur leaves related to the content of metabolites. The difference in absorption peaks between the groups became more expressed over time under the influence of drought. The pigment content in the leaf blades varied during the experiment, which indicates plant adaptation to stress. Preliminary results of the study can be used to expand knowledge about ways to evaluate the vital state of woody plants in the field.

Calculation of a parallaxpanoramogram in autostereoscopic systems with inconsistent monitor and lens raster parameters

N.V. Kondratiev, Yu.N. Ovechkis, A.I. Vinokur, D.A. Arsentiev

Accepted: 2024-07-11

Abstract

A significant disadvantage of the multi-point of view autostereoscopic method is a drop in image resolution with an increase in the number of points of view. An effective means of increasing the resolution is the use of an inclined lens raster and vertical encoding the colors of the point of view. Algorithmically simple coding is obtained at optimal tilt angles, the tangent of which is 1 divided by a multiple of 3 (1/3, 1/6, etc.). This requirement imposes significant restrictions on the coordination of the geometric parameters of the equipment – the display panel and the lens raster. The approaches to spatial color coding proposed in this article and the algorithms implementing them make it possible to significantly expand the possibilities of creating autostereoscopic displays. The experimental work carried out convincingly confirms the theoretical conclusions. The main practical result was the developed software that allows fine-tuning of the angle of inclination of the raster and calculating a multi-point of view parallaxpanoramogram for a specific set of equipment.

On the visualization of subattractor under mixed tidal forcing

Stepan Elistratov, Ivan But

Accepted: 2024-05-18

Abstract

One of the principle conditions of a wave attractor appearance is a periodic external forcing. Real forcing in natural basins caused by tidal interaction is more complex than a monochromatic which is usually used in internal wave attractors investigations. Multi-frequency forcing may lead to the multiple wave attractor formation, some of them may be of low energy, which affects their detection. In the article we simulate a mixed forcing for an internal wave attractor flow and visualize subattractor formed due to this forcing type using several methods, including Proper Orthogonal Decomposition. It is shown that the latter method reveals sub-attractor even in case of highly turbulent flow.

Visualization of turbulent wakes behind large particles

A.A. Mochalov, A.Yu. Varaksin

Accepted: 2024-05-17

Abstract

An attempt was made to visualize the flow formed in the wake of large particles moving in a downward turbulent airflow in the channel. The paper also considers the possibilities of reconstructing velocity fields behind a large particle from visual data. A diagram of the experimental setup is shown (geometry of the working area, auxiliary and main equipment). The PIV (Particle Image Velocimetry) system is briefly described. A technique for visualizing multiphase flow “gas – solid particles” is proposed. The original images of large particles (spheres) are shown. The results of the experimental determination of the characteristics of the wake vortex behind the rear critical point of a large particle are presented.