Neural Radiance Fields are a new technology regarding representation of digital scenes in a three-dimensional space. After an experimental study evaluating choices and technologies used in some of the most important implementations for the state of the art, this work has the objective of applying these new solutions in different practical environments, in particular in those related to the conservation of the Cultural Heritage. This to verify what these tool can represent and what might the best technical choices be, among those available, in obtaining the best representation result in the defined environment. In order to do this, some architectures will be tested with datasets and solution from the state of the art of Computer Graphics and Deep Learning. Afterwards, a report will be conducted confronting the best and most recent available architectures on three different datasets showing scenes to be synthesized related to the artistic and cultural heritage of the Marche region. Different approaches related to different conditions of the starting images acquisition will then be drawn up.
I campi di radianza neurale sono una nuova tecnologia per la rappresentazione di scene digitali in uno spazio tridimensionale. Questo lavoro vuole, a seguito di uno studio sperimentale che valuti scelte e tecnologie impiegate in alcune delle implementazioni più importanti per lo stato dell’arte, applicare queste nuove soluzioni in diversi ambienti pratici, in particolare in quelli legati alla conservazione del patrimonio culturale. Ciò per verificare quanto è rappresentabile da questi approcci e quali possano essere le migliori scelte tecniche, tra quelle a disposizione, per ottenere il più alto risultato possibile di rappresentazione contestualmente al Cultural Heritage. Per fare ciò, alcune architetture verranno testate con dataset e soluzioni allo stato dell’arte per Computer Graphics e Deep Learning. Dopodiché verrà effettuato un confronto tra le migliori e più recenti architetture disponibili su tre diversi dataset riportanti scene da sintetizzare legate al patrimonio artistico e culturale della regione Marche. Verranno quindi stilati diversi approcci relativi a diverse condizioni di acquisizione delle immagini di partenza.
Neural Radiance Fields: Architettura e Pipeline per il Cultural Heritage
CAPRARI, DAVID
2022/2023
Abstract
Neural Radiance Fields are a new technology regarding representation of digital scenes in a three-dimensional space. After an experimental study evaluating choices and technologies used in some of the most important implementations for the state of the art, this work has the objective of applying these new solutions in different practical environments, in particular in those related to the conservation of the Cultural Heritage. This to verify what these tool can represent and what might the best technical choices be, among those available, in obtaining the best representation result in the defined environment. In order to do this, some architectures will be tested with datasets and solution from the state of the art of Computer Graphics and Deep Learning. Afterwards, a report will be conducted confronting the best and most recent available architectures on three different datasets showing scenes to be synthesized related to the artistic and cultural heritage of the Marche region. Different approaches related to different conditions of the starting images acquisition will then be drawn up.File | Dimensione | Formato | |
---|---|---|---|
Caprari David - Tesi.pdf
accesso aperto
Descrizione: Documento di tesi di Caprari David.
Dimensione
36.82 MB
Formato
Adobe PDF
|
36.82 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/13677