Maxim Igorevich Sorokin

Dr. Maxim Igorevich Sorokin

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 June 2024 in Programming and Computer Software
Verified
Journal Article
Closed access
PDF file
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
DZDmitrii Zhdanov
Dmitrii Zhdanov
AZA. D. Zhdanov
A. D. Zhdanov
This paper presents an approach to 3D model reconstruction from point cloud data using modern neural network architectures. The method is based on PointNet++ and Transformer. PointNet++ plays a key role, providing efficient feature extraction and encoding of complex geometries in 3D scenes. This is achieved by recursively applying PointNet++ to nested partitions of the input point set in a metric space. Convex decomposition, which is an important step in the proposed approach, transforms complex 3D objects into sets of simpler convex shapes. This facilitates data processing and makes the reconstruction process more manageable. Then, Transformer trains the model on the extracted features, which enables the generation of high-quality reconstructions. It should be noted that Transformer is used only to determine positions of the walls and detect object boundaries. This combination of technologies allows us to achieve high accuracy of 3D model reconstruction. The main idea of the method is to segment the point cloud into small fragments, which are then reconstructed as polygon meshes. To restore missing points in point cloud data, a method based on the L1-median algorithm and local features of the point cloud is used. This approach can adapt to different geometric structures and correct topological connectivity errors. The proposed method is compared with several modern approaches, demonstrating its potential in various fields, including architecture, engineering, digitalization of cultural heritage, as well as augmented and mixed reality systems. This confirms its wide applicability and great perspectives for further development.

31 May 2020 in Programming and Computer Software
Verified
Journal Article
Closed access
PDF file
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
DZD. D. Zhdanov
D. D. Zhdanov
DZDmitrii Zhdanov
Dmitrii Zhdanov
IPIgor Potemin
Igor Potemin
NBN. N. Bogdanov
N. N. Bogdanov
One of the main problems of mixed reality devices is the lack of universal methods and algorithms for the visualization of virtual world objects in real space. The key point of natural perception of virtual objects in the real world is the creation of natural lighting conditions for virtual world objects by light sources located in the real world, i.e. the formation of natural glares on virtual objects and shadows cast by these objects in the real world. The paper proposes a method for adequately determining the position of the main light sources of the real world in mixed reality systems. Modern technologies that combine the capability of forming 2.5D images created by depth cameras and their subsequent computer processing using neural networks make it possible to identify real-world objects, recognize their shadows, and correctly restore the light sources that create these shadows. The results of the proposed method are presented, the accuracy of restoring the position of the light sources is estimated, and the visual difference between the image of the scene with the original light sources and the same scene with the restored parameters of the light sources is demonstrated.

26 May 2023 in Programming and Computer Software
Verified
Journal Article
Closed access
PDF file
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
DZDmitrii Zhdanov
Dmitrii Zhdanov
AZA. D. Zhdanov
A. D. Zhdanov
Mixed reality systems are a promising direction of research that opens up great opportunities for interaction with virtual objects in the real world. Like any promising direction, mixed reality has a number of unresolved problems. One of these problems is the synthesis of natural lighting conditions for virtual objects, including the correct light interaction of virtual objects with the real world. Since virtual and real objects are located in different spaces, it is difficult to ensure their correct interaction. To create digital copies of real-world objects, machine learning tools and neural network technologies are employed. These methods are successfully used in computer vision for space orientation and environment reconstruction. As a solution, it is proposed to transfer all objects into the same information space: virtual space. This makes it possible to solve most of the problems associated with visual discomfort caused by the unnatural light interaction of real and virtual objects. Thus, the basic idea of the method is to recognize physical objects from point clouds and replace these objects with virtual CAD models. In other words, it implies semantic analysis of a scene and classification of objects with their subsequent transformation into polygonal models. In this study, we use competitive neural network architectures, which can provide state-of-the-art results. The test experiments are carried out on Semantic3D, ScanNet, and S3DIS, which are currently the largest datasets with point clouds that represent indoor scenes. For semantic segmentation and classification of 3D point clouds, we use the PointNeXt architecture based on PointNet, as well as modern methods of data augmentation in the process of learning. For geometry reconstruction, the Soft Rasterizer differentiable rendering method and the Total3Understanding neural network are considered.

23 November 2024
Verified
Conference Paper
Closed access
PDF file
DZDmitry Zhdanov
Dmitry Zhdanov
VSVadim Sokolov
Vadim Sokolov
AZAndrei Zhdanov
Andrei Zhdanov
IPIgor S. Potemin
Igor S. Potemin
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
This paper presents a computer model of an optically transparent augmented reality system built using a configuration with two diffraction elements. A distinctive feature of this device is that the second diffraction element is a two-dimensional diffraction grating that combines functions of beam expansion and image extraction from the light guide. The technology for manufacturing two-dimensional diffraction gratings is presented. Formulas for calculating the direction of beams diffracted by the diffraction gratings, including the two-dimensional grating, are proposed. Two ray-tracing algorithms for image formation by optical augmented reality systems containing diffraction gratings, including two-dimensional ones, are proposed. Based on these algorithms, virtual prototyping of several augmented reality systems was conducted. Design features, advantages, and disadvantages of these systems were identified. Recommendations for the design and optimization of augmented reality systems using a dual diffraction element configuration are provided.

12 May 2025 in Programming and Computer Software
Verified
Journal Article
Closed access
PDF file
ALA. I. Lysykh
A. I. Lysykh
DZDmitrii Zhdanov
Dmitrii Zhdanov
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
The use of differentiable rendering methods is an up-to-date solution to the problem of geometry reconstruction from a set of RGB images without using expensive equipment. The disadvantage of this class of methods is the possible distortions of the geometry that arise during optimization and high computational complexity. Modern differentiable rendering methods calculate and use two types of gradients: silhouette gradients and normal gradients. Most distortions arising in geometry optimization are caused by modifications of parameters associated with silhouette gradients. The paper considers the possibility of increasing the efficiency of geometry reconstruction methods based on the use of differentiable rendering by dividing the reconstruction process into two stages: initialization and optimization. The first stage of reconstruction involves the creation of a visual Hull of the reconstructed object. This stage allows one to automate the process of selecting the original geometry and start the next stage under two conditions: the silhouettes of the object have already been reconstructed from all observation points and the topologies of the reconstructed and true objects are equivalent. The second stage comprises a geometry optimization cycle based on the fulfillment of the above conditions. This cycle consists of four steps: image rendering, loss function calculation, gradient calculation, and geometry optimization. Satisfying the condition of matching the contours of the original and reference geometry eliminates the need to use silhouette gradients. This solution significantly reduces the number of errors that occur during optimization, as well as reduces the computational complexity of the method by eliminating the calculation of the loss function, gradient calculation, and optimization of parameters associated with the silhouettes of objects. The testing and analysis of the results showed an increase in the accuracy of geometry reconstruction with a decrease in mesh resolution and a decrease in the total running time of the method in comparison with similar methods, as well as an up to two-fold increase in the speed of optimization steps.

1 January 2023
Verified
Conference Paper
Closed access
MSMaxim Igorevich Sorokin
Maxim Igorevich Sorokin
DZDmitrii Zhdanov
Dmitrii Zhdanov
AZAndrei Dmitrievich Zhdanov
Andrei Dmitrievich Zhdanov
Mixed reality systems create new forms of interaction between the physical and digital world by overlaying digital elements on the physical environment or creating virtual environments with physical elements. A key component for mixed reality are 3D scanning systems that capture the shape and texture of objects or scenes in the form of point clouds. For effective use of these data in mixed reality applications, they are transformed into polygonal meshes suitable for rendering, animation, and interaction. The proposed method includes the application of ResNet blocks for feature extraction, the use of an autoencoder to obtain latent 3D shapes, and the definition of wall and ceiling geometry using bounding boxes. The method allows obtaining a complete 3D model of the scene from a point cloud using deep learning and geometric analysis. In this work, two datasets were used for training and experiments: ShapeNet and ScanNet. These datasets represent a large and diverse collection of three-dimensional objects and scanned scenes with detailed annotation.