Obogatitev 3D upodobitve z globalnim osvetlitvenim modelom z uporabo generativnih nasprotniških nevronskih mrež

Zmazek, Marko

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Obogatitev 3D upodobitve z globalnim osvetlitvenim modelom z uporabo generativnih nasprotniških nevronskih mrež
Authors:	ID Zmazek, Marko (Author) ID Kohek, Štefan (Mentor) More about this mentor... ID Strnad, Damjan (Comentor)
Files:	MAG_Zmazek_Marko_2024.pdf (19,56 MB) MD5: 18A9B6A38FBCBD66C1B5362C1F5D546E
Language:	Slovenian
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	V računalniški grafiki je upodabljanje z globalnim osvetlitvenim modelom v realnem času še vedno aktualen izziv. Pri upodabljanju z globalnim osvetlitvenim modelom upodobitev izgleda bolj realistično, saj lahko dodamo učinke, kot sta odboj in lom svetlobe. V magistrskem delu smo implementirali in naučili generativno nasprotniško nevronsko mrežo, da obogati sliko 3D scene z globalnim osvetlitvenim modelom na podlagi slike 3D scene, upodobljene z lokalnim osvetlitvenim modelom, in dodatnih informacij o sceni, ki jih lahko hitro izračunamo že pri uporabi lokalnega osvetlitvenega modela. Ustrezno načrtovana in naučena nevronska mreža lahko na scenah, uporabljenih v fazi učenja, daje rezultate primerljive s klasičnimi metodami upodabljanja z globalnim osvetlitvenim modelom, kot je sledenje potem. Predvsem pa je lahko pri delovanju hitrejša, zato bi se lahko uporabljala za aplikacije v realnem času.
Keywords:	računalniška grafika, osvetlitev, generativne nasprotniške nevronske mreže, PyTorch, Blender
Place of publishing:	Maribor
Publisher:	[M. Zmazek]
Year of publishing:	2024
PID:	20.500.12556/DKUM-89729
UDC:	004.925:004.032.26(043.2)
COBISS.SI-ID:	218615555
Publication date in DKUM:	19.09.2024
Views:	0
Downloads:	58
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	07.08.2024

Secondary language

Language:	English
Title:	Enrichment of 3D rendering with a global illumination model using generative adversarial neural networks
Abstract:	In computer graphics, real-time rendering with a global illumination model is still a challenge. Rendering with a global illumination model makes the render look more realistic, as effects such as light reflection and refraction can be added. In this master thesis, we implemented and trained a generative adversarial neural network to enrich an image of a 3D scene with a global illumination model based on an image of a 3D scene rendered with a local illumination model and additional scene information, which can already be quickly computed when using the local illumination model. A properly designed and trained neural network can give us results comparable to classical global illumination model rendering methods, such as path tracing, on the scenes used in the learning phase. Above all, this approach can be faster and could therefore be used for real-time applications.
Keywords:	computer graphics, illumination, generative adversarial neural networks, PyTorch, Blender

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back