Interpolacija odsevnih sond za hitro upodabljanje odsevnih materialov

diplomsko delo

Katarina Gojković (Author), Matija Marolt (Mentor), Žiga Lesar (Co-mentor)

Abstract

Diplomsko delo obravnava problem upodabljanja odsevnih materialov v računalniški grafiki. Za reševanje tega problema smo razvili konvolucijski nevronski mreži, katerih cilj je bil ustvariti čim boljše slike okolice v določeni točki v sceni. Oba pristopa temeljita na pojavu prekomernega prilagajanja, zaradi česar smo morali vsako mrežo individualno učiti za specifične scene. Prva mreža je oblikovana tako, da na vhodu prejme x, y in z koordinate v sceni ter na izhodu generira sliko okolice v tej točki. Pri drugem pristopu pa smo sceno razdelili s triangulacijo in zajeli slike okolice v vseh ogliščih trikotnikov. Tako druga mreža na vhodu prejme tri slike okolice (zajete v ogliščih trikotnika) in uteži, ki odražajo oddaljenost točke v prostoru od teh oglišč, na izhodu pa poda napoved slike okolice v določeni točki v prostoru. Oba pristopa uspešno napovedujeta slike okolice v želenih točkah v prostoru, tudi če te niso bile del učne množice, vendar je njihova natančnost odvisna od kompleksnosti same scene. Obe metodi rešujeta problem ostrih prehodov med odsevnimi sondami ob premikanju po sceni in sta primerni za upodabljanje odsevov na premikajočih se objektih.

Keywords

odsevne sonde;interpolacija;konvolucijske nevronske mreže;univerzitetni študij;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2024
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[K. Gojković]
UDC:	004.92:004.032.26(043.2)
COBISS:	184620291
Views:	30
Downloads:	4
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Reflectance probe interpolation for fast rendering of reflective materials
Secondary abstract:	The thesis deals with the problem of rendering reflective materials in computer graphics. To solve this problem, we developed two convolutional neural networks, the goal of which was to create the best possible images of the surroundings at a certain point of the rendered scene. Both approaches are based on the phenomenon of overfitting, which made it necessary to train each network individually for specific scenes. The first network is designed in such a way that it receives x, y and z coordinates in the scene at the input and generates an image of the surroundings at that point at the output. In the second approach, we divided the scene by triangulation and captured images of the surrounding area in all vertices of the triangles. Thus, the second network at the input receives three images of the surroundings (captured in the vertices of the triangle) and weights that reflect the distance of a point from these vertices, and at the output it provides a prediction of the image of the surroundings at a certain point in space. Both approaches successfully predict images of the surroundings at desired points in space, even if these were not part of the training set, but their accuracy depends on the complexity of the scene itself. Both methods solve the problem of sharp transitions between reflection probes when moving around the scene and are thus suitable for rendering reflections on moving objects.
Secondary keywords:	computer graphics;reflectance probe;interpolation;convolutional neural network;computer and information science;diploma;Računalniška grafika;Nevronske mreže (računalništvo);Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000468
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	69 str.
ID:	22859745