Transformacija slika korištenjem generativnih difuzijskih modela

master's thesis

• Title (english): Image transformation using generative diffusion models
• Author: Lovro Glogar
• Mentor: Zoran Kalafatić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Siniša Šegvić (član povjerenstva)
• Committee member: Tomislav Hrkać (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-09, Croatia

Abstract

Ovaj rad istražuje primjenu difuzijskih generativnih modela za zadatke transformacije slike, posebno kolorizaciju i super-rezoluciju. Pruža sveobuhvatan pregled dubokog učenja i generativnih modela, detaljizirajući arhitekture kao što je U-Net i metode kao što su Palette i SR3. Eksperimenti koriste skup podataka ImageNet, pokazujući poboljšanja kvalitete slike pomoću ovih naprednih modela. Metrike kao što su PSNR, SSIM, Inception Score i FID koriste se za ocjenu uspješnosti. Studija naglašava... More snagu difuzijskih modela u stvaranju visokokvalitetnih slika i predlaže buduće smjerove istraživanja za rješavanje trenutnih ograničenja i poboljšanje izvedbe modela. Less

Abstract (english)

This work explores the application of generative diffusion models for image
transformation tasks, specifically colorization and super-resolution. It provides a
comprehensive overview of deep learning and generative models, detailing architectures
like U-Net and methods such as Palette and SR3. The experiments utilize the ImageNet
dataset, demonstrating significant improvements in image quality using these advanced
models. Metrics such as PSNR, SSIM, Inception Score, and FID are used for... More evaluation.
The study highlights the strengths of diffusion models in generating high-quality images
and suggests future research directions to address current limitations and enhance model
performance.
Less

• Keywords: duboko učenje; generativni modeli; transformacija slika; translacija slika-usliku; difuzijski generativni modeli
• Keywords (english): deep learning; generative models; image transformation; image-to-image transformation; diffusion generative models
• Language: croatian
• URN:NBN: urn:nbn:hr:168:311777
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2025-02-25 13:41:47