|
|
Visipedia - Embedding-driven Visual Feature Extraction and Learning
Jakeš, Jan ; Beran, Vítězslav (oponent) ; Zemčík, Pavel (vedoucí práce)
Multidimensional embedding is a powerful method of representing similarity measures among objects without the need for their explicit categorization. It has been increasingly used in recent years to annotate objects making an important part of the Visipedia project and its related work. This work explores the possibilities of learning from embedding-annotated images using their visual attributes and develops methods of predicting embedding coordinates for previously unseen images. It studies the relevant feature extraction and learning algorithms and describes the whole process of design and development of such a system using common machine learning approaches. The system is tested and evaluated with two different datasets and the performed experiments present the first results for a task of its kind.
|
|
| |
|
| |
|
|
Document Comparison Based on Visual Features
Milička, Martin ; Bartík, Vladimír (oponent) ; Burget, Radek (vedoucí práce)
The content of this thesis presents a design of the Web page comparison method that is based on visual features. At the beginning, the possible ways of the document comparison with regard to their use are described. The approaches concerning visual document comparison are presented in the next chapter. At first, the description is focused on the rendered image of page and then to a source code approach. This document is also focused on obtaining visual features from the source code. As a part of this thesis is a proposal of new approach for a document comparison based on visual features that uses structural document description. The proposal method is implemented in the application. One chapter also shows the results. The conclusion contains information for a future work.
|
|
|
Visipedia - Embedding-driven Visual Feature Extraction and Learning
Jakeš, Jan ; Beran, Vítězslav (oponent) ; Zemčík, Pavel (vedoucí práce)
Multidimensional embedding is a powerful method of representing similarity measures among objects without the need for their explicit categorization. It has been increasingly used in recent years to annotate objects making an important part of the Visipedia project and its related work. This work explores the possibilities of learning from embedding-annotated images using their visual attributes and develops methods of predicting embedding coordinates for previously unseen images. It studies the relevant feature extraction and learning algorithms and describes the whole process of design and development of such a system using common machine learning approaches. The system is tested and evaluated with two different datasets and the performed experiments present the first results for a task of its kind.
|
host ::
RSS.