Ανάκτηση εικόνων με βάση το περιεχόμενο με χρήση τεχνικών ημι-επιτηρούμενης μηχανικής μάθησης
Κατωπόδης, Βασίλειος Σ.
The main objective of this paper is the development of a software system which has the ability to retrieve images based on their content, in an automated way. The software tool extracts knowledge from the images’ features and the users’ ratings. This knowledge is used to rank the images in an automated and efficient manner. The learning algorithm of the software tool is based on the principles of the machine learning and more specifically on the principles of the semi-supervised learning. The system makes use of both labeled and unlabeled data for the training purposes. The software tool extracts the low-level description characterizations of the set of images that consists a part of the training set. Particularly extract MPEG-7 global visual descriptions from associated visual content of images. The extracted description characterizations measure the color and the texture of the given images. The software tool makes use of Transductive Support Vector Machine (TSVM) as the learning algorithm. The TSVMs extend SVMs and they follow the principles of transduction learning. While regular SVMs try to induce a general decision function for a learning task, this algorithm takes into account a particular test set. The software tool makes use of SVM light, an implementation of SVM, which proceeds by solving a sequence of optimization problems using a form of local search.