Full text
| |
| Author(s): |
Lucas Pascotti Valem
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | São José do Rio Preto. 2019-03-20. |
| Institution: | Universidade Estadual Paulista (Unesp). Instituto de Biociências Letras e Ciências Exatas. São José do Rio Preto |
| Defense date: | 2019-03-01 |
| Advisor: | Daniel Carlos Guimarães Pedronette |
| Abstract | |
Due to the evolution of technologies to store and share images, find effective methods to index and retrieve this type of information is indispensable. The CBIR (Content-Based Image Retrieval) systems are the main solution for image retrieval tasks. These systems consist in the use of different descriptors, supervised learning methods, and more recently unsupervised learning methods. However, accurately estimate the similarity between images is still a challenging task, mainly due to the well known semantic gap problem. As different ranked lists present different effectiveness results, an interesting methodology is to combine these ranked lists. This work proposes three different unsupervised methods for selecting and combining ranked lists. The approaches were evaluated in five different image collections and several descriptors. The methods achieved results comparable or superior to the state-of-the-art in most of the evaluated scenarios. (AU) | |
| FAPESP's process: | 17/02091-4 - Selection and combination of unsupervised learning Methdos for content-based image retrieval |
| Grantee: | Lucas Pascotti Valem |
| Support Opportunities: | Scholarships in Brazil - Master |