Exploring domain-informed and physics-guided learning in image-to-image translation

Pizzati, Fabio (2022) Exploring domain-informed and physics-guided learning in image-to-image translation, [Dissertation thesis], Alma Mater Studiorum Università di Bologna. Dottorato di ricerca in Automotive per una mobilità intelligente, 34 Ciclo. DOI 10.48676/unibo/amsdottorato/10499.
Documenti full-text disponibili:
[img] Documento PDF (English) - Richiede un lettore di PDF come Xpdf o Adobe Acrobat Reader
Disponibile con Licenza: Creative Commons Attribution Non-commercial No Derivatives 4.0 (CC BY-NC-ND 4.0) .
Download (124MB)

Abstract

Image-to-image (i2i) translation networks can generate fake images beneficial for many applications in augmented reality, computer graphics, and robotics. However, they require large scale datasets and high contextual understanding to be trained correctly. In this thesis, we propose strategies for solving these problems, improving performances of i2i translation networks by using domain- or physics-related priors. The thesis is divided into two parts. In Part I, we exploit human abstraction capabilities to identify existing relationships in images, thus defining domains that can be leveraged to improve data usage efficiency. We use additional domain-related information to train networks on web-crawled data, hallucinate scenarios unseen during training, and perform few-shot learning. In Part II, we instead rely on physics priors. First, we combine realistic physics-based rendering with generative networks to boost outputs realism and controllability. Then, we exploit naive physical guidance to drive a manifold reorganization, which allowed generating continuous conditions such as timelapses.

Abstract
Tipologia del documento
Tesi di dottorato
Autore
Pizzati, Fabio
Supervisore
Dottorato di ricerca
Ciclo
34
Coordinatore
Settore disciplinare
Settore concorsuale
Parole chiave
Image-to-image translation, GAN, domain bias, vision and physics, physics-guided learning, few-shot learning
URN:NBN
DOI
10.48676/unibo/amsdottorato/10499
Data di discussione
29 Novembre 2022
URI

Altri metadati

Statistica sui download

Gestione del documento: Visualizza la tesi

^