Title

Paste-and-Cut : Collective Image Localization and Classification for Real-Time Multi-Camera Object Detection

DGIST Authors

Young Eun Kang ; Hoon Sung Chwa ; Yeseong Kim

Advisor

좌훈승

Co-Advisor(s)

Yeseong Kim

Issued Date

2023

Awarded Date

2023-08-01

Citation

Young Eun Kang. (2023). Paste-and-Cut : Collective Image Localization and Classification for Real-Time Multi-Camera Object Detection. doi: 10.22677/THESIS.200000684351

Type

Thesis

Description

"Real-time systems"; "Object Detection"; "Neural Network"

Table Of Contents

Ⅰ.Introduction 1
Ⅱ.Related Work 5
Ⅲ.Target System and Motivation 7
3.1 Target System: DNN-based Object Detection 7
3.2 Feasibility of System Model 8
Ⅳ.System Design 12
4.1 System Goal and Overview 12
4.2 Merge-Canvas Size Decision Module 13
4.3 Localization Module: Paste 16
4.4 Classification Module: Cut 16
4.4.1 RoI Size Decision 17
4.4.2 Recursive Packing 18
Ⅴ. Evaluation 20
5.1 Experiment Setup 20
5.2 Experimental Results 21
5.2.1 Deadline Miss Ratio 21
5.2.2 Detection Accuracy 24
5.2.3 Execution time breakdown 27
Ⅵ. Conclusion 29
References 30

URI

http://hdl.handle.net/20.500.11750/46449
http://dgist.dcollection.net/common/orgView/200000684351

DOI

10.22677/THESIS.200000684351

Degree

Master

Department

Department of Electrical Engineering and Computer Science

Publisher

DGIST

Show Full Item Record