Uporaba umetne inteligence za preslikavo človeškega gibanja v 3D modele

magistrsko delo

Primož Stopar (Author), Milan Zorman (Mentor), Blaž Tomažič (Co-mentor)

Abstract

V magistrskem delu smo predstavili uporabo umetne inteligence za preslikavo človeškega gibanja v trodimenzionalni model na mobilnih napravah. Predstavili smo različne načine delovanja modelov umetne inteligence za ocenitev poze. Opisali smo nekaj že naučenih modelov umetne inteligence in ogrodij, s katerimi lahko modele umetne inteligence vključimo v igralni pogon Unity. Rešitev smo implementirali v pogonu Unity z ogrodjem MediaPipe in družino modelov umetne inteligence BlazePose. Ugotovili smo, da lahko že na povprečnih mobilnih napravah dosežemo skoraj realno časovno izvajanje. Prav tako menimo, da tehnologija še ni primerna za uporabo v zdravstvene namene, je pa primerna za uporabo v aplikacijah, namenjenih zabavi.

Keywords

vtičnik MediaPipe;model BlazePose;ocenitev poze;sledenje pozam;konvolucijske nevronske mreže;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[P. Stopar]
UDC:	004.8:004.932(043.2)
COBISS:	116087043
Views:	83
Downloads:	30
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Using artificial intelligence for mapping human motion into 3D models
Secondary abstract:	In the thesis we presented the use of artificial intelligence, to map human movement into three dimensional models on mobile devices. We have presented different designs of artificial intelligence models for pose estimation. We have described some already learned models of artificial intelligence and framework with which we can include models to the Unity engine. We implemented the solution in Unity engine, using MediaPipe framework and group of BlazePose models. We found out that we can achieve almost real-time performance on average mobile devices. We also believe that the technology is not yet suitable for use in medical purposes, but it can be used in applications intended for entertainment.
Secondary keywords:	MediaPipe;BlazePose;pose estimation;pose tracking;convolutional neural network;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Informatika in tehnologije komuniciranja
Pages:	1 spletni vir (1 datoteka PDF (X, 82 f.))
ID:	15752246