Pronalaženje teksta dubokim konvolucijskim modelima

undergraduate thesis

• Title (english): Text Localization With Deep Convolutional Models
• Author: Kristijan Fugošić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Petra Bevandić (član povjerenstva)
• Committee member: Marin Oršić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Korištenje dubokog učenja u metodama pronalaženja teksta, ali i u mnogim drugim problemima računalnog vida, dovelo je do značajnog napretka. Rezidualne mreže imaju puno bolje performanse od tradicionalnih neuronskih mreža pri većim dubinama te su se pokazale kao veoma korisne u svrhu pronalaženja teksta. Promatranjem rezultata natjecanja poput Robust Reading Competitiona vide se raznoliki pristupi pri rješavanju ovoga problema i napredak koji se ostvaruje kroz godine. Model za pronalaženje teksta EAST ima naglasak na jednostavnosti arhitekture i brzini izvođenja, te postiže veoma dobre rezultate.

Abstract (english)

The use of deep learning in text localization methods, but also in many other computer vision problems, has brought significant progress to the field. Residual networks perform much better than traditional neural networks at higher depths, and have proved to be very useful for text localization. By observing the results of competitions like Robust Reading Competition, we can see many different approaches to addressing this problem, as well as the progress that has been made over the years. A text-localization model EAST has an emphasis on performance and simple architecture, and achieves very good results.

• Keywords: računalni vid; duboko učenje; neuronske mreže; pronalaženje teksta; east
• Keywords (english): computer vision; deep learning; neural networks; text localization; east
• Language: croatian
• URN:NBN: urn:nbn:hr:168:484947
• 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: 2019-03-13 20:19:33