Implementation and Performance Analysis of Mobile Real-Time Frame Interpolation Network Using Deep Learning
- 주제(키워드) Deep learning , Video frame interpolation , Mobile , Optical flow , Real time
- 주제(DDC) 006.31
- 발행기관 아주대학교
- 지도교수 황원준
- 발행년도 2023
- 학위수여년월 2023. 8
- 학위명 석사
- 학과 및 전공 일반대학원 인공지능학과
- 실제URI http://www.dcollection.net/handler/ajou/000000032830
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
This research emphasizes the importance of constructing an efficient deep learning model that can perform real-time video frame interpolation in resource-constrained mobile environments, amidst the rapid advancements in the field of deep learning technology. The study proposes a lightweight network model and system for real-time video frame interpolation in mobile environments. By integrating intelligent data adjustment, lightweight CNN architecture, and distributed computing techniques, the model is designed to operate efficiently even with limited resources. The proposed lightweight network model contributes to the field of video frame interpolation by providing a lightweight solution tailored to mobile environments. It also opens up possibilities for various industries where efficient storage utilization and high frame rates are crucial. Furthermore, the research provides insights into lightweight techniques and distributed computing strategies that can be applied to other deep learning models in resource-constrained environments.
more목차
I. Introduction 1
II. Related Works 4
A. Video Frame Interpolation Techniques 4
B. Model Lightweighting Techniques 5
III. Proposed Method 8
A. Data pipeline 8
B. Convolutional Neural Network for Flow Estimation 10
C. Training 10
Ⅳ. Implementation Details 12
V. Experimental Results 14
A. Results on PC 14
B. Results on mobile 15
C. Network Comparison 15
Ⅵ. Conclusion 17
Ⅶ. Reference 18
