Title

Deep Learning-Based Privacy-Preserving Optical Image Captioning

Alternative Title

딥러닝 기반 프라이버시 보호 광학 이미지 캡셔닝

DGIST Authors

Martin Antoinette Deborah ; Inkyu Moon

Advisor

문인규

Issued Date

2026

Awarded Date

2026-02-01

Type

Thesis

Description

Deep Learning, Privacy-Preserving, Image Captioning, Partial Encryption, Photon Counting Imaging (PCI), Poisson–Multinomial Distribution–based PCI.

Table Of Contents

ABSTRACT i
List of Contents ii
List of Tables iv
List of Figures vi
1. INTRODUCTION 1
1.1. Motivations and Objectives 1
1.2. Overview 5
1.3. Contributions and Outline 6
2. PRIVACY-PRESERVING IMAGE CAPTIONING USING DOUBLE RANDOM PHASE ENCODING (DRPE) 8
2.1. Double Random Phase Encoding (DRPE) 8
2.2. Full DRPE-Based Image Captioning 10
2.2.1 Methodology 10
A. Overview 10
B. Encoder 11
C. Decoder 12
2.2.2. Experiments 15
A. Dataset 15
B. Implementation Details 17
C. Evaluation Metric 17
D. Results 20
2.3. Partial DRPE-Based Image Captioning 26
2.3.1. Methodology 26
A. Partial Encryption Using DRPE 26
B. Overview of Proposed Method 26
C. Dual-Stream Encoder 27
D. Transformer 29
2.3.2 Experiments 30
A. Dataset 30
B. Implementation Details 32
C. Ablation Study 32
1) Effect of Dual-Stream Encoder 32
2) Transformer Architecture Analysis 33
3) GloVe Embeddings 34
4) Fine-tuning Strategies on the Encoder 35
D. Quantitative Results and Visualization 37
E. Comparison with State of the Art 48
F. Additional Experiments 50
3. PRIVACY-PRESERVING OPTICAL IMAGE CAPTIONING USING PHOTON-LIMITED AND FEDERATED LEARNING 56
3.1. Methodology 56
3.1.1. Poisson-Multinomial Distribution-based Photon Counting Imaging (PMD-PCI) 56
3.1.2. Model Architecture 57
3.1.3 Federated Learning Framework 60
3.2. Experiments 64
3.2.1. Dataset 64
3.2.2. Implementation Details 64
3.2.3. Results 65
A. Centralized Training 67
B. Federated Learning 81
Ⅳ. CONCLUSION AND FUTURE WORK 96
4.1. Summary and Discussion 96
References 100
요 약 문 108

URI

https://scholar.dgist.ac.kr/handle/20.500.11750/59605
http://dgist.dcollection.net/common/orgView/200000942999

DOI

10.22677/THESIS.200000942999

Degree

Doctor

Department

Department of Robotics and Mechatronics Engineering

Publisher

DGIST

Show Full Item Record