Istovremena lokalizacija i segmentacija objekata

master's thesis

• Title (english): Simultaneous object localization and segmentation
• Author: Petra Marče
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Dejan Škvorc (član povjerenstva)
• Committee member: Leonardo Jelenković (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-19, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U okviru magistarskog rada proučavaju se duboke konvolucijske neuronske mreže za zadatke detekcije objekata i segmentacije slika. Metoda detekcije objekata SSD detaljno je opisana i isprobana na ispitnim skupovima PASCAL VOC2007, Cityscapes i KITTI. Za semantičku segmentaciju slika ispitnog skupa Cityscapes korištena je potpuno konvolucijska mreža s dilatiranim konvolucijama. Segmentacija instanci objekata ostvarena je kombiniranjem semantičke segmentacije i regresije vektora centroida objekta.

Abstract (english)

This master thesis studies approach to object detection and image segmentation using deep convolutional neural networks. The SSD object detection approach is thoroughly described and tested on PASCAL VOC2007, Cityscapes and KITTI datasets. For semantic segmentation on Cityscapes dataset, fully convolutional network with \textit{atrous} convolutions is used. Instance segmentation is achieved combining semantic segmentation and object centroid vector regression.

• Keywords: konvolucijske neuronske mreže; duboko učenje; klasifikacija; regresija; VGG16; detekcija objekata; semantička segmentacija; SSD; višezadaćno učenje; Cityscapes
• Keywords (english): convolutional neural networks; deep learning; classification; regression; VGG16; object detection; semantic segmentation; multi-task learning; Cityscapes
• Language: croatian
• URN:NBN: urn:nbn:hr:168:871136
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; 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 22:37:39