Please use this identifier to cite or link to this item:
Title: Applying deep learning for food image analysis
Author: Marrugat Torregrosa, Gerard
Director/Tutor: Radeva, Petia
Keywords: Visió per ordinador
Xarxes neuronals (Informàtica)
Tesis de màster
Sistemes classificadors (Intel·ligència artificial)
Aprenentatge automàtic
Computer vision
Neural networks (Computer science)
Masters theses
Learning classifier systems
Machine learning
Issue Date: 2-Sep-2019
Abstract: [en] Food is an important component in people’s daily life, examples of the previous assertion are the range of possible diets according to the animal or vegetal origin, the intolerance to some aliments and lately the increasing number of food pictures in social networks. Several computer vision approaches have been proposed for tackling food analysis problems, but few effort has been done in taking benefit of the hierarchical relation between elements in a food image; dish and ingredients. In this project the highly performing state of the art CNN method is adapted concatenating an ontology layer, a multidimensional layer which contains the relation between the elements, in order to help during the classification process. Different structures for the ontology have been tested to prove which relations have the most beneficial impact, and which are less relevant. Additionally to structure, the value of the elements that compound this hierarchical relation layer play an important role, therefore the experiments performed contained different weighted relations between the components. The ontology layer is built with the labels of the multiple task in the dataset used to train the model. At the end, the results obtained will be compared to a baseline model without the ontology layer and it will be appreciated how hierarchical relations between tasks benefits classification. Finally, the result will be a model which will be able to simultaneously predict two food-related tasks; dish and ingredients.
Note: Treballs finals del Màster de Fonaments de Ciència de Dades, Facultat de matemàtiques, Universitat de Barcelona, Any: 2019, Tutor: Petia Radeva
Appears in Collections:Programari - Treballs de l'alumnat
Màster Oficial - Fonaments de la Ciència de Dades

Files in This Item:
File Description SizeFormat 
food_analysis-master.zipCodi font12.33 MBzipView/Open
Github repository and blog entry.pdfRepositori Github297.23 kBAdobe PDFView/Open
160143.pdfMemòria2.08 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons