Dmitry Yudin

Dmitry Yudin
Moscow Institute of Physics and Technology | MIPT · Laboratory of Intelligent transport in the Center for cognitive modeling

PhD
MIPT, AIRI

About

34
Publications
12,775
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
109
Citations
Additional affiliations
April 2019 - March 2021
Moscow Institute of Physics and Technology
Position
  • Senior Researcher
September 2015 - April 2019
Belgorod State Technological University
Position
  • Professor (Associate)
October 2013 - April 2019
Belgorod State Technological University
Position
  • Researcher
Education
September 2015 - June 2017
Belgorod State Technological University
Field of study
  • Mechatronics and Robotics
September 2005 - June 2010
Belgorod State Technological University
Field of study
  • Automation of Techonological Processes and Productions

Publications

Publications (34)
Preprint
Full-text available
A neural network-based approach for hierarchical waste identification with poorly supervised object segmentation is described in the study. WaRP, a unique open labeled dataset, was created to train and evaluate suggested algorithms using industrial data from a waste recycling plant's conveyor. The dataset contains 28 different types of recyclable g...
Poster
Full-text available
This paper presents a simple yet efficient method called “Feature Map Flow, FMF” for 3D object detection and tracking, considering time-spatial feature map aggregation from different timesteps of deep neural model inference. Several versions of the FMF are proposed: from common concatenation to context-based feature map fusion and odometry usage fo...
Chapter
Generating depth maps using mono- or stereo- images is a topic of active research. This paper is dedicated to study of different methods of depth and disparity maps generating. It includes analysis of existing methods of depth maps generating and investigation and improvement of the real-time neural-network based method AnyNet. Our approach AnyNet-...
Article
The detection of dynamic and static obstacles is a key task for the navigation of autonomous ground vehicles. The article presents a new algorithm for generating an occupancy map of the surrounding space from noisy point clouds obtained from one or several stereo cameras. The camera images are segmented by the proposed deep neural network FCN-ResNe...
Chapter
Full-text available
In navigation systems for unmanned vehicles, an important task is fusion of the pose estimations (odometry and localization) obtained from different sensors: cameras, LiDARs, wheel encoders, inertial measurement modules, etc. To solve this task, it is necessary to know the covariance matrices for each of the odometry sources, which characterize the...
Preprint
Full-text available
In this paper, we present a real-time 3D detection approach considering time-spatial feature map aggregation from different time steps of deep neural model inference (named feature map flow, FMF). Proposed approach improves the quality of 3D detection center-based baseline and provides real-time performance on the nuScenes and Waymo benchmark. Code...
Article
Full-text available
The paper is devoted to the task of multiple objects tracking and segmentation on monocular video, which was obtained by the camera of unmanned ground vehicle. The authors investigate various architectures of deep neural networks for this task solution. Special attention is paid to deep models providing inference in real time. The authors proposed...
Article
Full-text available
Paper discusses modern methods for localization of mobile ground robot in outdoor environment. Special attention is payed to lidar based methods which allow to do precise simultaneous localization and mapping (SLAM) independently to lighting conditions. Modular approach to lidar-based localization is proposed. It consists of three modules: map reco...
Conference Paper
Full-text available
The paper describes usage of deep neural networks for flat roof defect segmentation on aerial images. Such architectures as U-Net, DeepLabV3+ and HRNet+OCR are studied for recognition five categories of roof defects: “hollows”, “swelling”, “folds”, “patches” and “breaks”. Paper introduces RoofD dataset containing 6400 image pairs: aerial photos and...
Conference Paper
Full-text available
Navigation of unmanned vehicle especially using orthophoto is a topic of active research. This paper is dedicated to study of different methods of ortho-photo-based localization methods. For this task new dataset was created. It consists of pairs of ground level and bird's eye view images collected on vehicle test site of the technology contest Up...
Chapter
Full-text available
The paper describes usage of deep neural networks based on ResNet and Xception architectures for recognition of age and gender of imbalanced dataset of face images. Described dataset collection process from open sources. Training sample contains more than 210000 images. Testing sample have more 1700 special selected face images with different ages...
Chapter
Full-text available
The paper describes usage of modern deep neural network architectures such as ResNet, DenseNet and Xception for the classification of facial expressions on color and grayscale images. Each image may contain one of eight facial expression categories: “Neutral”, “Happiness”, “Sadness”, “Surprise”, “Fear”, “Disgust”, “Anger”, “Contempt”. As the datase...
Article
Full-text available
In the last years, deep learning and reinforcement learning methods have significantly improved mobile robots in such fields as perception, navigation, and planning. But there are still gaps in applying these methods to real robots due to the low computational efficiency of recent neural network architectures and their poor adaptability to robotic...
Article
Full-text available
Among a number of problems in the behavior planning of an unmanned vehicle the central one is movement in difficult areas. In particular, such areas are intersections at which direct interaction with other road agents takes place. In our work, we offer a new approach to train of the intelligent agent that simulates the behavior of an unmanned vehic...
Poster
Full-text available
Poster on the paper "Detection of Big Animals on Images with Road Scenes using Deep Learning"
Poster
Full-text available
Poster on the paper "Traffic Sign Recognition on Video Sequence Using Deep Neural Networks and Matching Algorithm". In this work we study different methods of object detection for the problem of traffic sign recognition on video sequence. We study still image detection approach and found that it may be improved with Seq-Bbox Matching. However, if t...
Conference Paper
Full-text available
The paper analyzes data sets containing images with labeled traffic signs, as well as modern approaches for their detection and classification on images of urban scenes. Particular attention is paid to the recognition of Russian types of traffic signs. Various modern architectures of deep neural networks for the simultaneous object detection and cl...
Conference Paper
Full-text available
The recognition of big animals on the images with road scenes has received little attention in modern research. There are very few specialized data sets for this task. Popular open data sets contain many images of big animals, but the most part of them is not correspond to road scenes that is necessary for on-board vision systems of unmanned vehicl...
Article
Full-text available
The paper considers the task solution of detection on two-dimensional images not only face, but head of a human regardless of the turn to the observer. Such task is also complicated by the fact that the image receiving at the input of the recognition algorithm may be noisy or captured in low light conditions. The minimum size of a person’s head in...
Article
In the construction and technical expertise of buildings, inspection is an important and crucial part in the conclusions about the technical condition of the object. It is carried out visually and instrumentally with the interpretation of results by subjective expert methods. The current performance standard of construction and technical inspection...
Chapter
Full-text available
The paper describes usage of deep neural network architectures such as VGG, ResNet and InceptionV3 for the classification of small images. Each image may contain one of four vehicle pose categories or background. An iterative procedure for training a neural network is proposed, which allows us to quickly tune the network using wrongly classified im...
Chapter
Full-text available
The article describes the application of various machine learning methods for the analysis of images obtained from a video camera with the purpose of detection its partial or total visibility loss. Computational experiments were performed on a data set containing more than 6800 images. Support vector machine, categorical boosting and simplified mod...
Conference Paper
Full-text available
The paper considers usage of fine-tuning of the deep neural network ensemble for recognition of 60 event types in the set of 60,000 images from WIDER database. The applied ensemble consists of two deep convolutional neural networks (CNN) using the GoogLeNet architecture, previously trained on other image bases: ImageNet and Places. Separately the a...
Article
Full-text available
The article presents information on software for roof defects recognition on aerial photographs, made with air drones. An areal image segmentation mechanism is described. It allows detecting roof defects – unsmoothness that causes water stagnation after rain. It is shown that HSV-transformation approach allows quick detection of stagnation areas, t...
Article
Full-text available
This article discusses the approach to developing modular software for real-time control of an industrial construction 3D printer. The proposed structure of a two-level software solution is implemented for a robotic system that moves in a Cartesian coordinate system with multi-axis interpolation. An algorithm for the formation and analysis of a pat...
Conference Paper
Full-text available
This article describes the development of control system of robotic complex for constructions and buildings printing. Based on analysis of existing approaches authors propose the solution for control system development for 3D-printer with gantry-type construction, which can move on rails. Fully functional model of the control system was created by...
Article
Full-text available
Event recognition on images plays important role in the context of events search and review, archiving and storage of photos, advertising, media. Paper presents a classification system based on texture features (Multi-Level Histogram of Color Multi-Scale Local Binary Pattern and Local Derivative Pattern) and support vector machine (SVM). For classi...
Conference Paper
Full-text available
The article describes the task of green control of the energy-intensive object on the example of the rotary cement kiln. It represents the control system model on basis of firing process statistical data, obtained by conventional sensors and vision system. To determine the "input-output" correlation three methods are applied: the classical regressi...
Conference Paper
Full-text available
Article analyses modern machine vision-based approaches of the rotary kiln monitoring and control which reduces energy consumption and improve clinker quality. Article describes fuzzy advising control unit developed by the authors. It based on sintering zone state assessment, the kiln rotation period and the relative change of the exhaust gases tem...
Article
Full-text available
Article describes developed hardware-software complex of machine vision system for assessment of firing process parameters in rotary cement kilns. Authors developed method of firing process image recognition, allowing automatically real-time assessment of the firing process in three parameters: dust, state of the material and state of the torch. It...
Article
Full-text available
The paper describes machine vision system (MVS) developed by authors, which identifies artificial landmarks on images from a video camera with pan-tilt mechanism and allows to calculate the deviation of the robot from the set course. Lines limiting robot track and tags in the form of two-dimensional barcodes were selected as artificial landmarks. D...

Network

Cited By

Projects

Projects (3)
Archived project
- development and implementation of energy-effective deep learning algorithms for road scene recognition (object classification, detection and segmentation). Industrial partner: Scientific-Design Burеau of Computing Systems (SDB CS)
Project
Development of methods and tools for constructing a robust autonomous control system of vehicle and its groups that ensure their safe goal achievement
Archived project
Development of deep learning algorithms for recognition of traffic conditions on images and their application in unmanned vehicles