Document type
Master thesisPublication date
Publication license
Please use this identifier to cite or link to this item: https://hdl.handle.net/2445/180444
Attention Guided Image-to-Image translation of Food Images Using GANs
Journal Title
Authors
Director/Tutor
Journal ISSN
Volume Title
Related resource
Abstract
[en] Nowadays, with the help of the novel machine learning models, in particular Generative Adversarial Networks, we are able to generate synthetic media, which look absolutely realistic and at the same time authentic. Still, the food image-to-image translation remains a challenging problem that is very unexplored. Due to the complexity of food images the state of the art results are noisy and slow in convergence. In our work, we explore how adding attention to the image-to-image translation on food data can produce more realistic synthetic images and speed up the convergence of the algorithm. Furthermore, we present extensive analysis of GANs for food image synthesis and discuss several possible improvements over the base methodology sharing our insights on this problem.
The source code that has been used to produce the results in this project can be found in our GitLab repository: https://gitlab.com/deep-food-ub/food-gan.
Description
Treballs finals del Màster de Fonaments de Ciència de Dades, Facultat de matemàtiques, Universitat de Barcelona, Any: 2020, Tutor: Petia Radeva i Javier Ródenas Cumplido
Subject (English)
Citation
Citation
MLADENOVA, Mariya and TONCHEV, Petar. Attention Guided Image-to-Image translation of Food Images Using GANs. [consulted: 6 of June of 2026]. Available at: https://hdl.handle.net/2445/180444