|
|
Application for Recognition of People by Face
Svoboda, Jakub ; Orság, Filip (referee) ; Goldmann, Tomáš (advisor)
Person identification has in the recent years gained notoriety as one of the most powerful ways of extracting information from image data. This thesis is focused on the task of human identification from facial photographs. To solve this task, we employ algorithms based on neural networks, which produce more robust results than traditional algorithms. In this thesis, we studied the common approaches for solving this problem and based on the gathered knowledge we created an architecture of a neural network trained to tackle the task of human identification and verification based on facial photographs. We have then further improved the model architecture and the training process by performing various experiments and observing the results. The final model has reached an accuracy comparable to other state-of-the-art models. Furthermore, we created a desktop application to demonstrate the results visually and to enable easier manipulation with the identity database. The knowledge gathered in this thesis can be used for improvements of current identification models or models modified for solving similar tasks.
|
|
|
Mobile Application to Control the Diet of Patients on Dialysis
Pavlacký, Ondřej ; Kolář, Dušan (referee) ; Goldmann, Tomáš (advisor)
The goal of this thesis is to create an intuitive mobile application for Android operation system which will help dialysis patients by displaying approximate values of monitored substances in their body and replace classic paper records. Thanks to this mobile application patients are going to have better overview of consumed foodstuffs a help them stick to their diet. When inserting the foodstuff patients will select from the database of foodstuffs and the application will calculate values that the food contains. Patients can also expand the database with their own food. Application will also register periods between dialysis and from the length of the period, patients' weight and their diuresis calculate maximal values of monitored substances (water, potassium and phosphorus). Patients will also see values from passed periods, which will we shown in statistics. Data will be persistently saved in local and remote database. Application is developed on Xamarin platform.
|
|
|
Algorithm for Head Comparison in Non-Standard Views
Wysoglad, Jaromír ; Goldmann, Tomáš (referee) ; Drahanský, Martin (advisor)
The goal of this work is to create an algorithm for human head comparison. The algorithm is able to compare heads in a lot of different positions, but the heads, that are being compared, must be in the same position. At first the algorithm uses some freely available detectors for detecting heads and head parts. Then a histogram of oriented gradients is computed for each part of each head and by comparing them the algorithm finds out the dissimilarity of the heads. From the testing set of 30 pictures the algorithm is able to successfully detect heads on 26 pictures. Every picture was compared with 5 other pictures, with one of them containing a head of the same person. If I don't count the 4 pictures, where the algorithm wasn't able to detect the head and 2 pictures, which should have been assigned to these pictures. The algorithm successfully determined, that the head of the same person is the most similar on 18 pictures. On 5 pictures the head of the same person was determined as the 3rd most similar and on one of the pictures the algorithm failed completely and determined the head of the same person to be the least similar. The algorithm is successful with head comparison in different positions on most of the pictures.
|
|
|
Face Liveness Detection Using a 2D Camera
Valo, Ondrej ; Drahanský, Martin (referee) ; Goldmann, Tomáš (advisor)
Facial recognition is one of the most socially accepted forms of biometric recognition. The recent availability of highly accurate and efficient face recognition algorithms leaves vulnerability to presentation attacks as a major challenge for face recognition solutions. This work deals with the explanation of the issues related to the detection of facial liveliness, which will help to understand the various possibilities of attack and their relationship to existing solutions. And the implementation of an algorithm that recognizes the liveliness of the face based on videos.
|
|
|
Intelligent Thermal Camera with Intruder Detection
Mysza, Róbert ; Drahanský, Martin (referee) ; Goldmann, Tomáš (advisor)
Thermovision technology is becoming more and more widespread in monitoring and security applications. This bachelor thesis deals with its usage for object detection and describes methods and algorithms capable to do this. I also studied factors, which effects the human skin temperature. The design of the device, which sends images to the server and is able to detect an intruder by his temperature is part of this thesis. The last chapter contains evaluation of tests for detection reliability.
|
|
|
Person Detection in Video
Marek, Lukáš ; Goldmann, Tomáš (referee) ; Dyk, Tomáš (advisor)
This bachelor thesis deals with the implementation of an application for detecting people in video footage. Both the application and the detection are implemented in Python. TKinter library is used to create the application. The OpenCV library and the YOLOv4 detection algorithm, which runs on a CUDA backend but can also run on the CPU, are used for detection.
|
|
|
Simulation of an ARM Processor for the Education of Programming in Assembler
Ondryáš, Ondřej ; Goldmann, Tomáš (referee) ; Orság, Filip (advisor)
This thesis aims to implement a didactic tool for simulation of an Arm-based processor integrated into the Visual Studio Code editor. The tool facilitates learning about the machine-level programming of these processors. It implements a service that provides an assembler and a simulator for the A32 instruction set. The service is built using the Unicorn emulation framework and other open-source tools. The editor extension uses the service to add support for the development and debugging of programs written in the assembly language. It shows descriptions of used instructions and helps the programmer understand their function. When debugging, it enables stepping through the code and provides various views of the state of the simulated processor, its registers and memory. The solution can be used in the Advanced Assembly Languages course at FIT BUT. It could be further improved in the future to support other architectures and provide an easy learning environment in other courses related to machine-level programming.
|
|
|
Thermal Imaging System for Measuring Body Temperature
Mysza, Róbert ; Kolář, Dušan (referee) ; Goldmann, Tomáš (advisor)
The COVID-19 pandemic brought increased need for measuring human temperature. This thesis deals with solution of using low-cost thermal camera module FLIR Lepton 3.5 for measuring human forehead temperature and examines the real usability of this in terms of an accuracy. In the beginning, I describe various methods of measuring temperature and factors, which can influence the measurement. Subsequently, I examine various factors influencing human body and surface temperature. As of result of thesis is full design and implementation of thermal image system for temperature measurement, which I tested in different environmental conditions and compared its the precision against medical contactless infrared thermometer.
|
|
|
Detection and Recognition of Gun in a Scene
Stuchlík, David ; Goldmann, Tomáš (referee) ; Drahanský, Martin (advisor)
The aim of the diploma thesis is to design an algorithm for detection and recognition of the type of gun in the image. Firstly, the existing methods and techniques for detecting the various objects are briefly introduced in the text of the thesis however, the methods are primarily focused on guns. Next, the basics of neural networks are briefly outlined, followed by an overview of the most common detectors for deep neural networks. The second half of the thesis is devoted to the implementation of an application for generating images based on a 3D model of a gun, the creation of a data file and learning of a neural network. Finally, the results obtained, which clearly indicate that in order to cover a huge variation of real weapons, is necessary to generate a large amount of training data based on many different 3D models, are briefly summarized in the conclusion of the thesis.
|
|
|
Estimation of the Position of the Feet in a Image
Havlíček, Lukáš ; Orság, Filip (referee) ; Goldmann, Tomáš (advisor)
This thesis deals with the design and implementation of solution for estimating the position of the feet in the image, focusing on cases where the feet are covered in the image. The area of anthropometry, neural networks, solution design and implementation and subsequent experiments are described here. The solution uses own convolutional network, implemented using the Keras interface with the creation of own dataset with a focus on covered legs. In the experiments, other points on the body are used to increase accuracy.
|
guest ::
RSS feed.