|
|
Object detection for video surveillance using the SSD approach
Dobranský, Marek ; Lokoč, Jakub (advisor) ; Božovský, Petr (referee)
The surveillance cameras serve various purposes ranging from security to traffic monitoring and marketing. However, with the increasing quantity of utilized cameras, manual video monitoring has become too laborious. In re- cent years, a lot of development in artificial intelligence has been focused on processing the video data automatically and then outputting the desired no- tifications and statistics. This thesis studies the state-of-the-art deep learning models for object detection in a surveillance video and takes an in-depth look at SSD architecture. We aim to enhance the performance of SSD by updating its underlying feature extraction network. We propose to replace the initially used VGG model by a selection of modern ResNet, Xception and NASNet classifica- tion networks. The experiments show that the ResNet50 model offers the best trade-off between speed and precision, while significantly outperforming VGG. With a series of modifications, we improved the Xception model to match the ResNet performance. On top of the architecture-based improvements, we ana- lyze the relationship between SSD and a number of detected classes and their selection. We also designed and implemented a new detector with the use of temporal context provided by the video frames. This detector delivers enhanced precision while...
|
|
|
Artificial neural networks for pattern recognition
Kukačka, Marek ; Mrázová, Iveta (advisor) ; Božovský, Petr (referee)
This work describes the advantages and disadvantages of using neural networks for pattern recognition. Several neural network models are described and their use for pattern recognition is demonstrated. Standard multi-layered perceptron model is compared to a more sophisticated convolutional network model. A new network model is introduced, which is inspired by the convolutional networks and aimed at rectifying some of their shortcomings. The work describes results of tests performed with the described network model on the problem of recognizing hand-written digits.
|
|
|
Deep neural networks and their implementation
Vojt, Ján ; Mrázová, Iveta (advisor) ; Božovský, Petr (referee)
Deep neural networks represent an effective and universal model capable of solving a wide variety of tasks. This thesis is focused on three different types of deep neural networks - the multilayer perceptron, the convolutional neural network, and the deep belief network. All of the discussed network models are implemented on parallel hardware, and thoroughly tested for various choices of the network architecture and its parameters. The implemented system is accompanied by a detailed documentation of the architectural decisions and proposed optimizations. The efficiency of the implemented framework is confirmed by the results of the performed tests. A significant part of this thesis represents also additional testing of other existing frameworks which support deep neural networks. This comparison indicates superior performance to the tested rival frameworks of multilayer perceptrons and convolutional neural networks. The deep belief network implementation performs slightly better for RBM layers with up to 1000 hidden neurons, but has a noticeably inferior performance for more robust RBM layers when compared to the tested rival framework. Powered by TCPDF (www.tcpdf.org)
|
|
|
Neural network architectures for mobile devices
Georgiev, Georgi Stoyanov ; Mrázová, Iveta (advisor) ; Božovský, Petr (referee)
Designing effective methods for image classification and real-time object detection is one of the most well-known problems of the present. A series of convolutional neural networks has been designed in order to solve these tasks. Neural networks created spe- cifically for mobile devices are among the fastest ones. In this work we focus primarily on the MobileNetV2 and EfficientNetB0 models. We present their structure and compare them with one another. We research several algorithms designed to automatically build new neural network models as well. An essential part of the convolutional network design process is the optimization of their structure. We outline sensitivity analysis methods which help us observe how network inputs influence its outputs, and pruning methods designed to remove redundant neurons. In the end we demonstrate an example usage of the EfficientNetB0 model in a mobile appliaction created to classify cars. 1
|
|
|
Data generator
Pečimúth, Andrej ; Kopecký, Michal (advisor) ; Božovský, Petr (referee)
The work focuses on the design of a tabular data generator. Our solu- tion reads the structure of input data coming from different sources. This schema can be further modified using a graphical environment. Parame- terizable generators are associated with each column. The generators and their parameters are automatically selected by the system so that the output records resemble the input ones. The output set of records respects integrity constraints. Multiple output formats are supported. The work combines the advantages of existing web and native solutions. We bring functional- ity previously only available in native applications to the web environment. In addition, we have solved the problem of circular dependencies between tables. 1
|
|
|
Object detection for video surveillance using the SSD approach
Dobranský, Marek ; Lokoč, Jakub (advisor) ; Božovský, Petr (referee)
The surveillance cameras serve various purposes ranging from security to traffic monitoring and marketing. However, with the increasing quantity of utilized cameras, manual video monitoring has become too laborious. In re- cent years, a lot of development in artificial intelligence has been focused on processing the video data automatically and then outputting the desired no- tifications and statistics. This thesis studies the state-of-the-art deep learning models for object detection in a surveillance video and takes an in-depth look at SSD architecture. We aim to enhance the performance of SSD by updating its underlying feature extraction network. We propose to replace the initially used VGG model by a selection of modern ResNet, Xception and NASNet classifica- tion networks. The experiments show that the ResNet50 model offers the best trade-off between speed and precision, while significantly outperforming VGG. With a series of modifications, we improved the Xception model to match the ResNet performance. On top of the architecture-based improvements, we ana- lyze the relationship between SSD and a number of detected classes and their selection. We also designed and implemented a new detector with the use of temporal context provided by the video frames. This detector delivers enhanced precision while...
|
|
|
Deep neural networks and their implementation
Vojt, Ján ; Mrázová, Iveta (advisor) ; Božovský, Petr (referee)
Deep neural networks represent an effective and universal model capable of solving a wide variety of tasks. This thesis is focused on three different types of deep neural networks - the multilayer perceptron, the convolutional neural network, and the deep belief network. All of the discussed network models are implemented on parallel hardware, and thoroughly tested for various choices of the network architecture and its parameters. The implemented system is accompanied by a detailed documentation of the architectural decisions and proposed optimizations. The efficiency of the implemented framework is confirmed by the results of the performed tests. A significant part of this thesis represents also additional testing of other existing frameworks which support deep neural networks. This comparison indicates superior performance to the tested rival frameworks of multilayer perceptrons and convolutional neural networks. The deep belief network implementation performs slightly better for RBM layers with up to 1000 hidden neurons, but has a noticeably inferior performance for more robust RBM layers when compared to the tested rival framework. Powered by TCPDF (www.tcpdf.org)
|
|
|
Artificial neural networks and reinforcement learning
Havránek, Vojtěch ; Božovský, Petr (referee) ; Mrázová, Iveta (advisor)
When solving complex machine learning tasks, it is often more practical to let the agent find an adequate solution by itself using e.g. reinforcement learning rather than trying to specify a solution in detail. The only information required for reinforcement learning is a reward that gives the agent reinforcement about the desirability of his actions. Our experiments suggest that good results can be achieved by reinforcement learning with online learning neural networks. The functionality of such neural network may be further extended by allowing it to model the environment and/or by providing it with recurrent connections. In this thesis, we show that for a given network predicting the reward, it is NP-complete to find the agent action that maximizes this reward. We describe three neural network models, one of them being an original modification of Sutton's TD(¸) algorithm that extends its domain to non-Markovian environments. All three models were thoroughly tested with our predator-prey simulator. The most powerful of them, the modified TD(¸) was then applied to control of a real mobile robot. Simultaneously, we have discussed the principles of rewarding the agents, the biological plausibility of the algorithms, the importance of the exploration capabilities and general bounds of reinforcement learning....
|
|
|
Artificial neural networks for pattern recognition
Kukačka, Marek ; Božovský, Petr (referee) ; Mrázová, Iveta (advisor)
This work describes the advantages and disadvantages of using neural networks for pattern recognition. Several neural network models are described and their use for pattern recognition is demonstrated. Standard multi-layered perceptron model is compared to a more sophisticated convolutional network model. A new network model is introduced, which is inspired by the convolutional networks and aimed at rectifying some of their shortcomings. The work describes results of tests performed with the described network model on the problem of recognizing hand-written digits.
|
guest ::
RSS feed.