Izgubno stiskanje slik z variacijskim avtokodirnikom

magistrsko delo

Žiga Leber (Author), Damjan Strnad (Mentor), Štefan Kohek (Co-mentor)

Abstract

V magistrskem delu preučujemo izgubno stiskanje slik z uporabo variacijskega avtokodirnika. Implementirali smo njegovo povratno različico, imenovano konvolucijski DRAW, ki v vlogi kodirnika in dekodirnika uporablja nevronsko mrežo LSTM. Za implementacijo smo uporabili jezik Python in knjižnico PyTorch. Delovanje algoritma smo testirali na podatkovni zbirki CIFAR-10 ter rezultate primerjali z metodo JPEG. Ugotovili smo, da so rezultati primerljivi v smislu kakovosti rekonstrukcije.

Keywords

stiskanje slik;nevronske mreže;variacijski avtokodirniki;konvolucijski DRAW;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2018
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	Ž. Leber
UDC:	004.627:004.932(043.2)
COBISS:	21878294
Views:	645
Downloads:	121
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Lossy image compression using variational autoencoder
Secondary abstract:	In this thesis we study lossy image compression using variational autoencoder. We implemented its recurrent variant called convolutional DRAW, which uses a LSTM neural network in the role of the encoder and the decoder. The implementation was done in Python using the PyTorch library. The performance was tested the CIFAR-10 dataset and the results compared to the JPEG compression method. We determined that the results are comparable in reconstruction quality.
Secondary keywords:	image compression;neural netwoks;variational autoencoder;convolutional DRAW;
URN:	URN:SI:UM:
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	XIII, 40 str.
ID:	10958139