Weakly-supervised semantic segmentation of hand-written equations

master's thesis

• Title: Slabo nadzirana semantička segmentacija rukom pisanih jednadžbi
• Author: Josip Milić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Slobodan Ribarić (član povjerenstva)
• Committee member: Jan Šnajder (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-17, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Semantička segmentacija je važan zadatak računalnog vida. Cilj tog zadatka je odrediti semantički razred svakog pojedinog piksela slike strojno naučenim modelom. Kao i u drugim problemima računalnog vida, najbolji rezultati postižu se strogo nadziranim učenjem. Međutim, veliki nedostatak tog pristupa je potreba za ručnim označavanjem svakog pojedinog piksela u skupu slika za učenje. Kako bismo smanjili potrebu za tim skupim procesom, razmatramo mogućnost slabo nadziranog učenja segmentacijskih modela. U ovom pristupu na temelju anotacije na razini okvira u skupu podataka za učenje dobiva se anotacija na razini piksela znakova rukom pisanih jednadžbi u skupu podataka za ispitivanje.

Abstract (english)

Semantic segmentation is an important task of computer vision. The goal of this task is to determine the semantic class of each individual pixel in a machine-learned model. As in other computer vision problems, the best results are achieved by strictly controlled learning. However, the major disadvantage of this approach is the need to manually mark each pixel in a learning image set. To reduce the need for this expensive process, we consider the possibility of poorly supervised learning of segmentation models. In this approach, based on the bounding box level annotation in the train dataset, annotation is obtained at the level of character pixels of handwritten equations in the test dataset.

• Keywords: slabo; nadzirana; semantička; segmentacija; rukom; pisana; jednadžba; računalni; vid
• Keywords (english): weakly; supervised; semantic; segmentation; handwritten; equation; computer; vision
• Language: croatian
• URN:NBN: urn:nbn:hr:168:902575
• Study programme: Title: Računarstvo; Study programme type: sveučilišni; Study level: diplomski; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-03-13 23:09:34