Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Deep Learning Quantification: Extracting Quantitative Information from Images using Convolutional Autoencoders
KTH, School of Information and Communication Technology (ICT).
2017 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

We solve the problem of retrieving quantitative information from input images, focusing on the domain of food ingredients. The goal is an application that takes one or multiple pictures of ingredients as input, and is able to provide a variety of suggestions about them, such as weight and calories. To solve this problem we build a fully-convolutional pipeline of neural networks. We use convolutional encoder-decoder architectures to solve the subproblems of classifying the objects in a picture, filtering them from the background, and reconstructing their 3D shapes. We then devise a method to obtain the volume of the objects, by comparing their 3D shapes with the reconstruction of a predefined reference object. Additionally, to solve the sub-problem of not having enough labeled data, we study the utilization of 3D synthetic models, by creating a novel synthetic dataset. Thus, we provide some experiments to study the power of networks trained with synthetically generated images. We provide a functioning pipeline that solves the aforementioned quantification task. Finally, we also propose a novel multitasking architecture.

Place, publisher, year, edition, pages
2017. , 87 p.
Series
TRITA-ICT-EX, 2017:112
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-211719OAI: oai:DiVA.org:kth-211719DiVA: diva2:1130550
Subject / course
Computer Science
Educational program
Master of Science in Engineering - Information and Communication Technology
Examiners
Available from: 2017-08-10 Created: 2017-08-10 Last updated: 2018-01-13Bibliographically approved

Open Access in DiVA

No full text

By organisation
School of Information and Communication Technology (ICT)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 305 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf