메뉴 건너뛰기
.. 내서재 .. 알림
소속 기관/학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
Md Atikuzzaman (Kyung Hee University) Sung-Ho Bae (Kyung Hee University)
저널정보
한국방송·미디어공학회 한국방송미디어공학회 학술발표대회 논문집 한국방송·미디어공학회 2024 하계학술대회
발행연도
2024.6
수록면
1,178 - 1,182 (5page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
Recent years have seen a surge in the popularity of the Denoising Diffusion Probabilistic Model (DDPM) for various computer vision tasks. While DDPM offers advantages over other techniques such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), it is hampered by slow sampling speeds, which limit its practical applications. To address this limitation, we propose a unique framework that processes images in grayscale during the diffusion process, thereby reducing computational costs. Additionally, we integrate a colorization technique to restore color, transforming the grayscale diffusion model generated images back into vibrant RGB images. By combining the superior sampling quality of DDPM with the efficiency of grayscale processing and subsequent colorization, our method aims to overcome the sampling speed barrier, making DDPM more practical for a wide range of computer vision tasks. This approach offers an optimal compromise between image quality and computational speed, paving the way for fast sampling and large-scale image generation ventures. Our method can be used as a plug-and-play module with any existing diffusion model. Notably, our method performs three times faster than classic DDPM for the CelebA dataset.

목차

Abstract
1. Introduction
2. Related Works
3. Method
5. Experimental Setups
6. Results and Analysis
6. Conclusion
References

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

최근 본 자료

전체보기

댓글(0)

0

UCI(KEPA) : I410-ECN-151-24-02-090142906