|
|
Automatic Photography Categorization
Veľas, Martin ; Beran, Vítězslav (referee) ; Španěl, Michal (advisor)
This thesis deals with content based automatic photo categorization. The aim of the work is to create an application, which is would be able to achieve sufficient precision and computation speed of categorization. Basic solution involves detection of interesting points, extraction of feature vectors, creation of visual codebook by clustering, using k-means algorithm and representing visual codebook by k-dimensional tree. Photography is represented by bag of words - histogram of presence of visual words in a particular photo. Support vector machines (SVM) was used in role of classifier. Afterwards the basic solution is enhanced by dividing picture into cells, which are processed separately, computing color correlograms for advanced image description, extraction of feature vectors in opponent color space and soft assignment of visual words to extracted feature vectors. The end of this thesis concerns to experiments of of above mentioned techniques and evaluation of the results of image categorization on their usage.
|
|
|
Navigation Using Deep Convolutional Networks
Skácel, Dalibor ; Veľas, Martin (referee) ; Hradiš, Michal (advisor)
In this thesis I deal with the problem of navigation and autonomous driving using convolutional neural networks. I focus on the main approaches utilizing sensory inputs described in literature and the theory of neural networks, imitation and reinforcement learning. I also discuss the tools and methods applicable to driving systems. I created two deep learning models for autonomous driving in simulated environment. These models use the Dataset Aggregation and Deep Deterministic Policy Gradient algorithms. I tested the created models in the TORCS car racing simulator and compared the result with available sources.
|
|
|
Robot Localization Using OpenStreet Map
Rajnoch, Zdeněk ; Veľas, Martin (referee) ; Rozman, Jaroslav (advisor)
Goal of this thesis is localization of mobile robot in OpenStreet map segment. Robot IMU, odometry and compass sensors are used for trajectory reconstruction, which is compared to reference GPS trajectory. Extended Monte Carlo localization and clusterization are used for robot localization. Software is implemented in C++ with ROS middleware.
|
|
|
Traffic Signs Detection and Localisation
Kudláč, Ondrej ; Španěl, Michal (referee) ; Veľas, Martin (advisor)
This thesis aims to design the traffic signs detection and localization system using RGB image and 3D LiDAR data leveraging the the existing solutions. Traffic sign detection is based on the shape analysis. Then, the LIDAR data are used for the localization of previously detected signs. The created solution consists of two main components: the detector and locator, each able to operate independently.
|
|
|
Automatic Content-Based Image Categorization
Němec, Ladislav ; Španěl, Michal (referee) ; Veľas, Martin (advisor)
This thesis deals with automatic content-based image classification. The main goal of this work is implementation of application which is able to perform this task automatically. The solution consists of variable system using local image features extraction and visual vocabulary built by k-means method. Bag Of Words representation is used as a global feature describing each image. Support Vector Machines - the final component of this system - perform the classification based on this representation. In the last chapter, the results of this experimental system are presented.
|
|
|
Calibration of Robotic Workspace
Uhlíř, Jan ; Veľas, Martin (referee) ; Kapinus, Michal (advisor)
This work is concerned by the issue of calibrating a robotic workplace, including the localization of a calibration object for the purpose of calibrating a 2D or 3D camera, a robotic arm and a scene of robotic workplace. At first, the problems related to the calibration of the aforementioned elements were studied. Further, an analysis of suitable methods for performing these calibrations was performed. The result of this work is application of ROS robotic system providing methods for three different types of calibration programs, whose functionality is experimentally verified at the end of this work.
|
|
|
Planar Object Measurement in Image
Mlýnek, Přemysl ; Veľas, Martin (referee) ; Beran, Vítězslav (advisor)
The aim of this work is to measure furniture doors using image processing and calculating the price of new doors from measured dimensions. The work is solved using OpenCV library and Python programming language. The core of the work is based on the FloodFill algorithm and Hough transforms. The graphical user interface is solved using the PyQt library. When obtaining the measurements of the furniture door, the following procedure is used - acquisition of a data set, imagine preprocessing, object segmentation, object classification, measurement of individual objects, output of measurement results and output price calculation. I have created a solution in the form of a desktop application that accepts an image at the input and outputs the measured dimensions along with the price of the new door. I managed to achieve an average deviation of 6 mm from real dimensions. This work has helped me understand the basics of image processing. An ordinary user of the application will be able to estimate the price of new furniture doors just based on a photo.
|
|
|
3D Model of a Room Using Kinect
Zemek, Martin ; Veľas, Martin (referee) ; Španěl, Michal (advisor)
This thesis is about finding significant planar surfaces in point cloud and their conversion to polygons. Which is important step for making a 3D model of a room. Input is point cloud, which was recorded by Kinect v2 sensor. Tool for capturing one snapshot from Kinect is included in this thesis. For recording more detailed point cloud is needed external program. Some of the programs are mentioned further in this thesis. For plane detection is used RANSAC. Inliers are divided using Euclidean Cluster Extraction. These clusters are converted to polygon using convex or concave hull. Application is capable of working with one snapshot or bigger point cloud assembled by registration of particular snapshots and detect primary and secondary planar surfaces. For the largest plane points can be prepared for creation of a texture and dimensions of this plan can be printed in CLI.
|
|
|
Moving Object Detection in the Environment of Mobile Robot
Dorotovič, Viktor ; Beran, Vítězslav (referee) ; Veľas, Martin (advisor)
This work's aim is movement detection in the environment of a robot, that may move itself. A 2D occupancy grid representation is used, containing only the currently visible environment, without filtering in time. Motion detection is based on a grid-based particle filter introduced by Tanzmeister et al. in Grid-based Mapping and Tracking in Dynamic Environments using a Uniform Evidential Environment Representation. The system was implemented in the Robot Operating System, which allows for re-use of modules which the solution is composed of. The KITTI Visual Odometry dataset was chosen as a source~of LiDAR data for experiments, along with ground-truth pose information. Ground segmentation based on Loopy Belief Propagation was used to filter the point clouds. The implemeted motion detector is able to distiguish between static and dynamic vehicles in this dataset. Further tests in a simulated environment have shown some shortcomings in the detection of large continuous moving objects.
|
|
|
Deep Learning for Image Classification
Ziková, Jana ; Veľas, Martin (referee) ; Hradiš, Michal (advisor)
This bachelor thesis deals with electronic commerce website products classification using product's photographs. For this purpose we use already implemented models of deep convolutional neural networks. Tho goal of this theses is to design experiments that will lead to the best possible results in product images classification.
|
guest ::
RSS feed.