DEVELOPMENT OF REAL-TIME OBJECT DETECTION ON SMARTPHONE BY IMPLEMENTING COMPUTATION OFFLOADING
Smartphones has become a tool that greatly supports the needs of people around the world and demand that applications with intensive computing can be done on smartphones. One of them is object detection using deep learning. In dealing with this problem, the mobile cloud computing paradigm emerged...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/70699 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Smartphones has become a tool that greatly supports the needs of people around the
world and demand that applications with intensive computing can be done on
smartphones. One of them is object detection using deep learning. In dealing with
this problem, the mobile cloud computing paradigm emerged with a technique
called computation offloading. The technique aims to extend the capabilities of
smartphones with remote resources in the cloud.
This research will develop the YOLOv3 object detection implemented with
computation offloading for smartphones to utilize cloudlet computing power. The
smartphone will function to take pictures, send and receive images to the cloudlet,
and display the results of object detection to the user. Cloudlet will process the
object detection for the received image and send the results back to the smartphone.
The communication protocol used for computation offloading is TCP, which
focuses on ensuring the success of sending and receiving data, as well as the
integrity of the data.
Based on the test results, the total processing time of object detection with
computation offloading is 6.3x faster than using local computing on smartphones,
but there is a decrease in average of confidence score by 4% on detected object. In
addition, processing with computing offloading can still be improved by replacing
the 2.4 Ghz to 5 Ghz wifi connection with the total processing time becoming 10%
faster, and the confidence score increasing by 1.5%. |
|---|