|
|
Active learning for Bayesian neural networks in image classification
Belák, Michal ; Šabata, Tomáš (advisor) ; Vomlelová, Marta (referee)
In the past few years, complex neural networks have achieved state of the art results in image classification. However, training these models requires large amounts of labelled data. Whereas unlabelled images are often readily available in large quantities, obtaining l abels takes considerable human effort. Active learning reduces the required labelling effort by selecting the most informative instances to label. The most popular active learning query strategy framework, uncertainty sampling, uses uncertainty estimates of the model being trained to select instances for labelling. However, modern classification neural networks often do not provide good uncertainty estimates. Baye sian neural networks model uncertainties over model parameters, which can be used to obtain uncertainties over model predictions. Exact Bayesian inference is intractable for neural networks, however several approximate methods have been proposed. We experiment with three such methods using various uncertainty sampling active learning query strategies.
|
|
|
Impact of color models on performance of convolutional neural networks
Šimunský, Martin ; Doležel, Petr (referee) ; Škrabánek, Pavel (advisor)
Current knowledge about impact of colour models on performance of convolutional neural network is investigated in the first part of this thesis. The experiment based on obtained knowledge is conducted in the second part. Six colour models HSV, CIE 1931 XYZ, CIE 1976 L*a*b*, YIQ a YCbCr and deep convolutional neural network ResNet-101 are used. RGB colour model achieved the highest classification accuracy, whereas HSV color model has the lowest accuracy in this experiment.
|
|
|
Detecting the Occurrence of Objects in a Video
Šamánek, Jan ; Orság, Filip (referee) ; Goldmann, Tomáš (advisor)
This bachelor thesis deals with detection of objects in videos by using primarily convolution neural networks and creating simple user interface, which allows user to choose classification model and use it to analyze video or train given model on own dataset. First part is dedicated to description of machine learning and neural networks. After that follows the section about image description and image classification using machine learning algorithms and data augmentation. Last part deals with describtion of own design of neural network and user interface and describing achieved results.
|
|
|
Vehicle Make and Model Recognition in Image
Buchta, Martin ; Špaňhel, Jakub (referee) ; Herout, Adam (advisor)
This thesis deals with classification of a car model from an image. It describes several methods, such as convolutional neural networks, methods limited to the fron/rear view and methods using 3D CAD models. From these approaches it chooses convolutional neural networks, which it further deals with. The work contains a description of the individual layers of which such a network consists. The practical part describes the procedure by which the classifier, that has an accuracy of 80.7\,\%, was created. A dataset containing 1\,034 photos was created to verify functionality. The work further experiments with different architectures and evaluates their accuracy. The work contains a program which, thanks to the car detector, finds the vehicle in the video and marks it with a square and a description of the car model in the given video.
|
|
|
Design of learning and equipment module using AI on Raspberry PI and Intel Movidius platform
Macko, Tomáš ; Richter, Miloslav (referee) ; Janáková, Ilona (advisor)
This bachelors thesis describes the process of implementing trained neural network model to AI accelerator - Intel Movidius. The first chapter is about machine learning and computer vision theory. The second chapter describes the options which can be chosen for programming of convolutional neural networks as programming language or related libraries which suit the most. The third and fourth chapters are highly connected. They describe the whole process of hardware installation and troubleshooting of software issues during installation. The next chapter shows previews of images, which are used as data input for neural network. Next pages describe used scripts and models of neural networks which were created from scratch. The last chapters are all about measured datas during the training or testing of neural networks and its evaluation.
|
|
|
Anatomy based landmark detection in brain CT scans
Krajčiová, Alexandra ; Harabiš, Vratislav (referee) ; Jakubíček, Roman (advisor)
Manual detection of anatomical landmarks from head CT (Computed Tomography) scans is time-consuming task prone to observer errors. In addition, the accuracy of the detection correlates with image quality. The aim of this work is to create an algorithm that will perform automatic detection of anatomical landmarks. These landmarks can be later used to form radiological lines, which finds its application in CT scanning. SVM (Support Vector Machines) and HOG (Histograms of Oriented Gradients) features was chosen for anatomical landmark detection. The achieved results, possibilities of further progress and improvement of detection are summarized in the conclusion.
|
|
|
Retinal biometry with low resolution images
Smrčková, Markéta ; Odstrčilík, Jan (referee) ; Kolář, Radim (advisor)
This thesis attempts to find an alternative method for biometric identification using retinal images. First part is focused on the introduction to biometrics, human eye anatomy and methods used for retinal biometry. The essence of neural networks and deep learning methods is described as it will be used practically. In the last part of the thesis a chosen identification algorithm and its implementation is described and the results are presented.
|
|
|
Deep Neural Networks Approximation
Stodůlka, Martin ; Mrázek, Vojtěch (referee) ; Vaverka, Filip (advisor)
The goal of this work is to find out the impact of approximated computing on accuracy of deep neural network, specifically neural networks for image classification. A version of framework Caffe called Ristretto-caffe was chosen for neural network implementation, which was extended for the use of approximated operations. Approximated computing was used for multiplication in forward pass for convolution. Approximated components from Evoapproxlib were chosen for this work.
|
|
|
The effect of the background and dataset size on training of neural networks for image classification
Mikulec, Vojtěch ; Kolařík, Martin (referee) ; Rajnoha, Martin (advisor)
This bachelor thesis deals with the impact of background and database size on training of neural networks for image classification. The work describes techniques of image processing using convolutional neural networks and the influence of background (noise) and database size on training. The work proposes methods which can be used to achieve faster and more accurate training process of convolutional neural networks. A binary classification of Labeled Faces in the Wild dataset is selected where the background is modified with color change or cropping for each experiment. The size of dataset is crucial for training convolutional neural networks, there are experiments with the size of training set in this work, which simulate a real problem with the lack of data when training convolutional neural networks for image classification.
|
|
|
Food classification using deep neural networks
Kuvik, Michal ; Přinosil, Jiří (referee) ; Burget, Radim (advisor)
The aim of this thesis is to study problems of deep convolutional neural networks and the connected classification of images and to experiment with the architecture of particular network with the aim to get the most accurate results on the selected dataset. The thesis is divided into two parts, the first part theoretically outlines the properties and structure of neural networks and briefly introduces selected networks. The second part deals with experiments with this network, such as the impact of data augmentation, batch size and the impact of dropout layers on the accuracy of the network. Subsequently, all results are compared and discussed with the best result achieved an accuracy of 86, 44% on test data.
|
guest ::
