Zulkalnain, Mohd Asyraf (2019) Optimization Of Intruder Detection Algorithm Using Raspberry Pi Platform. Project Report. UTeM, Melaka, Malaysia. (Submitted)
|
Text
Optimization of intruder detection algorithm using raspberry Pi platform.pdf Download (439kB) | Preview |
Abstract
This project involves the use of Convolutional Neural Network (CNN) architecture to develop an intruder detection algorithm. Basically, the intruder detection algorithm involves image classification task to classify input image into intruder and non-intruder. Recently, CNN have shown to have great accuracy on image classification application. Thus, this project is based on LeNet and MobileNet models, where both of the models were trained to classify intruder and non-intruder images. After various DNN models have been trained, the best model in terms of performance were ported from Personal Computer (PC) to Raspberry Pi 3 Model B+ in the deployment phase. MobileNet have shown to have high accuracy while maintaining low model complexity or number of operations. MobileNet v3 was chosen to be ported to Raspberry Pi because it has proven to have highest accuracy when tested with testing images of intruder and non-intruder. When the porting is complete, the performance in terms of accuracy and speed of MobileNet v3 in Raspberry Pi and PC was compared and evaluated. The average speed of MobileNet v3 when running on Raspberry Pi is 2.855 fps while on PC is 6.263 fps. The speed in terms of frames per second (fps) of MobileNet v3 when executed on Raspberry Pi was 54.41% slower compare to PC. This conclude that DNN model runs slower on Raspberry Pi compared to when executed on PC.
Item Type: | Final Year Project (Project Report) |
---|---|
Uncontrolled Keywords: | Image processing, Computer algorithms, Neural networks (Computer science) |
Subjects: | T Technology > TA Engineering (General). Civil engineering (General) |
Divisions: | Library > Final Year Project > FKEKK |
Depositing User: | Sabariah Ismail |
Date Deposited: | 25 Jun 2020 08:05 |
Last Modified: | 18 Aug 2020 06:10 |
URI: | http://digitalcollection.utem.edu.my/id/eprint/24387 |
Actions (login required)
![]() |
View Item |