Please use this identifier to cite or link to this item:
Title: Where am I eating? Image-based food menu recognition
Author: Valdivia Arriaza, Marc
Director/Tutor: Bolaños Solà, Marc
Keywords: Processament digital d'imatges
Visió per ordinador
Treballs de fi de grau
Reconeixement de formes (Informàtica)
Cartes (Restauració)
Aprenentatge automàtic
Digital image processing
Computer vision
Computer software
Pattern recognition systems
Bachelor's thesis
Machine learning
Issue Date: 27-Jun-2018
Abstract: [en] Food has become a very important aspect of our social activities. Since social networks and websites like Yelp appeared, their users have started uploading photos of their meals to the Internet. This factor leads to the development of food analysis models and food recognition. We propose a model to recognize the meal appearing in a picture from a list of menu items (candidates dishes). Which could serve for the recognize the selected meal in a restaurant. The system presented in this thesis does not need to train a new model for every new restaurant in a real case scenario. It learns to identify the components of an image and the relationship that they have with the name of the meal. The system introduced in this work computes the similarity between an image and a text sequence, which represents the name of the dish. The pictures are encoded using a combination of Convolutional Neural Networks to reduce the input image. While, the text is converted to a single vector applying a Long Short Term Memory network. These two vectors are compared and optimized using a similarity function. The similarity-based output is then used as a ranking algorithm for finding the most probable item in a menu list. According to the Ranking Loss metric, the results obtained by the model improve the baseline by a 15%.
Note: Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2018, Director: Marc Bolaños Solà
Appears in Collections:Treballs Finals de Grau (TFG) - Enginyeria Informàtica
Programari - Treballs de l'alumnat

Files in This Item:
File Description SizeFormat 
codi_font.zipCodi font380.39 kBzipView/Open
memoria.pdfMemòria4.61 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons