Vgrajen sistem za procesiranje video tokov na osnovi strojno pospešenih nevronskih mrež

magistrsko delo

Primož Kosi (Author), Iztok Kramberger (Mentor), Marko Kos (Co-mentor)

Abstract

V magistrskem delu je opisan vgrajen sistem za procesiranje video tokov na osnovi strojno pospešenih nevronskih mrež. Video tok se zajema s kamero ali preko HDMI vmesnika. Sistem je grajen za prepoznavanje predmetov v realnem času. Vhodni video tok se procesira s strojnim pospeševalnikom CoreVectorBlox, ki izvaja nevronske mreže. Nadzoruje ga mehka procesna enota RISC-V. Nevronske mreže so vključene v obliki binarnih podatkov. Povezave posameznih podsistemov so izvedene z AMBA AXI, AHB in APB vodili. Izhodni video tok se prikazuje na zaslonu preko HDMI povezave. Opisani so posamezni deli in delovanje celotnega sistema. V nalogi so predstavljene meritve porabe moči in učinkovitosti prepoznave predmetov.

Keywords

mikrokrmilniki;programirljiva logična vezja;procesiranje video tokov;strojno pospeševanje;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. Kosi]
UDC:	004.312.466(043.2)
COBISS:	129780483
Views:	98
Downloads:	14
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Embedded system for video processing based on hardware accelerated neural networks
Secondary abstract:	The master's thesis describes an integrated video stream processing system based on hardware accelerated neural networks. The video stream is captured by a camera or via HDMI interface. The system is built for real-time object recognition. The input video stream is processed by a CoreVectorBlox hardware accelerator, that implements neural networks. It is controlled by a RISC-V soft processing unit. The neural networks are included in the form of binary large objects. Connections of the individual subsystems are implemented by AMBA AXI, AHB and APB buses. The output video stream is displayed on the screen via HDMI connection. The individual components and the operation of the complete system are described. Measurements of power consumption and object recognition performance are presented in the thesis.
Secondary keywords:	Microcontrolers;Programmable Logic Device;Video strem processing;hardware acceleration;neural networks.;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Elektrotehnika
Pages:	1 spletni vir (1 datoteka PDF (X, 54 f.))
ID:	16263974