Raspoznavanje slika primjenom slojeva pažnje

undergraduate thesis

• Title (english): Image recognition with self-attention layers
• Author: Bruno Ćorić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Petra Bevandić (član povjerenstva)
• Committee member: Matej Grcić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Raspoznavanje slika važno je područje računalnog vida kojim dominiraju duboke konvolucijske mreže godinama, no nova arhitektura mreža temeljena na slojevima pažnje postaje sve popularnija u dubokom učenju osobito na području obrade prirodnog jezika. U ovom radu istražen je mehanizam pažnje i arhitekture koje ga koriste, istraženo je kako implementirati takve arhitekture za raspoznavanje slika te koje su prednosti i nedostatci. Implementirana je jedna takva arhitektura te su nad njom provedeni eksperimenti kako radi takva mreža te usporedbe s drugim popularnim mrežama u području raspoznavanja slika u računalnom vidu.

Abstract (english)

Image recognition is an important area of computer vision dominated by deep convolutional networks for years but the new attention-based network architectures are becoming increasingly popular in deep learning, especially in the field of natural language processing. In this paper, the mechanism of attention and the architectures that use it are investigated, we investigate how to implement such architectures for image recognition and what are the advantages and disadvantages. One such architecture has been implemented and experiments on how such a network works and comparisons with other popular networks in the field of computer image recognition have been performed.

• Keywords: pažnja; računalni vid; raspoznavanje slika; transformer
• Keywords (english): attention; computer vision; image classification; transformer
• Language: croatian
• URN:NBN: urn:nbn:hr:168:978794
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-14 20:15:04