Department of Engineering Master's Defence: Sverrir Bjarnason

Can neural networks detect previously unseen objects without human interference?

  • 15.1.2021, 10:00 - 12:00, Háskólinn í Reykjavík

Þann 15.janúar nk mun Sverrir Bjarnason verja meistaraverkefnið sitt sem ber titilinn Can neural networks detect previously unseen objects without human interference?

Dagsetning: 15.janúar frá 10:00-12:00
Stofa: M106
Nemandi: Sverrir Bjarnason
Leiðbeinandi: Torfi Þórhallson, Lektor, Háskólinn í Reykjavík
Prófdómari: Yngvi Björnsson, Prófessor, Háskólinn í Reykjavík

Titill: Can neural networks detect previously unseen objects without human interference?

Abstract: Automation brings with it many previously neglected problems as the unfathomable processing power of the human mind creates the illusion that detecting and picking up objects is, in some way, an easy task. Distinguishing objects from many other identical objects, often only in one color, is not as easy as one might think. The task at hand is to reduce the time it takes to train a neural network to recognize each object and annotate thousands of images to do so. Going through the process of finding a good data set to use in a series of experiments, this thesis aims to answer the question if there is a possibility to use a minimal dataset to train a neural network that can detect a much wider array of objects. Through incremental and batch learning procedure, experiments are done with different methods of training and detecting objects. From the results of these experiments, it can be concluded that there is a possibility that a network can be made to detect a large variety of objects by only training the network on a portion of the catalog at hand.



Vinsamlegast athugið að á viðburðum Háskólans í Reykjavík (HR) eru teknar ljósmyndir og myndbönd sem notuð eru í markaðsstarfi HR. Hægt er að nálgast frekari upplýsingar á ru.is eða með því að senda tölvupóst á netfangið: personuvernd@ru.is
//
Please note that at events hosted at Reykjavik University (RU), photographs and videos are taken which might be used for RU marketing purposes. Read more about this on our ru.is or send an e-mail: personuvernd@ru.is