지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학술저널
- 저자정보
- 발행연도
- 2023.12
- 수록면
- 875 - 887 (13page)
- DOI
- 10.5909/JBE.2023.28.7.875
이용수
초록· 키워드
오류제보하기
There have been many works on single image super-resolution (SISR) using convolutional neural networks (CNNs), researching on network architecture, loss function, applications, etc. However, very few have studied the modification or adaptation of the convolution operation, which is the fundamental element of CNN. In most CNN-based methods, the filter weights do not change at the inference phase, i.e., filter parameters are fixed regardless of the input and its regional characteristics. We note that this conventional approach is parameter-efficient but may not be optimal in performance due to its inflexibility to regionally different input statistics. To tackle this problem, we propose a novel convolution operation named Adaptive Convolution, which has content-specific characteristics. The proposed method adaptively adjusts filter weights according to the regional characteristics of the input with the help of an attention mechanism. We also introduce a kernel fragmentation method, which enables the efficient implementation of the Adaptive Convolution. We embed our new convolutional layer into several well-known SR networks and show that it enhances their performances while requiring a small number of additional parameters. Also, our method can be used along with other attentions that manipulate the features, further increasing the performance.
목차
Abstract
Ⅰ. Introduction
Ⅱ. Related Works
Ⅲ. Adaptive Convolution
Ⅳ. Experimental Results
Ⅴ. Conclusion
References
참고문헌 (50)
참고문헌 신청
Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super-resolution: Dataset and study. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 126 135, 2017. doi: https://doi.org/10.1109/cvp
Namhyuk Ahn, Byungkon Kang, and Kyung-Ah Sohn. Fast, accurate, and lightweight super-resolution with cascading residual network. In Proceedings of the European Conference on Computer Vision (ECCV), pages 252 268, 2018. doi: https://doi.org/10.1007/978-3-0
Marco Bevilacqua, Aline Roumy, Christine Guillemot, and Marie Line Alberi-Morel. Low-complexity single-image super-resolution based on nonnegative neighbor embedding. 2012. doi: https://doi.org/10.5244/c.26.135
Hong Chang, Dit-Yan Yeung, and Yimin Xiong. Super-resolution through neighbor embedding. In Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., volume 1, pages I I. IEEE, 2004. doi: https:
Tao Dai, Jianrui Cai, Yongbing Zhang, Shu-Tao Xia, and Lei Zhang. Second-order attention network for single image super-resolution. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 11065 11074, 2019. doi: https://doi
Namhyuk Ahn, Byungkon Kang, and Kyung-Ah Sohn. Fast, accurate, and lightweight super-resolution with cascading residual network. In Proceedings of the European Conference on Computer Vision (ECCV), pages 252 268, 2018. doi: https://doi.org/10.1007/978-3-0
Marco Bevilacqua, Aline Roumy, Christine Guillemot, and Marie Line Alberi-Morel. Low-complexity single-image super-resolution based on nonnegative neighbor embedding. 2012. doi: https://doi.org/10.5244/c.26.135
Hong Chang, Dit-Yan Yeung, and Yimin Xiong. Super-resolution through neighbor embedding. In Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., volume 1, pages I I. IEEE, 2004. doi: https:
Tao Dai, Jianrui Cai, Yongbing Zhang, Shu-Tao Xia, and Lei Zhang. Second-order attention network for single image super-resolution. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 11065 11074, 2019. doi: https://doi
함께 읽어보면 좋을 논문
논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!
-
FFT 적용을 통한 Convolution 연산속도 향상에 관한 연구
멀티미디어학회논문지
2021 .11
-
Convolution Neural Network Front-end / Back-end Design
대한전자공학회 학술대회
2020 .11
-
고해상도 영상을 위한 합성곱 신경망의 필터 크기에 대한 화질 비교
대한전자공학회 학술대회
2018 .06
-
New Approach to Optimize the Size of Convolution Mask in Convolutional Neural Networks
한국컴퓨터정보학회논문지
2016 .01
-
Deep Learning을 위한 GPGPU 기반 Convolution 가속기 구현
전기전자학회논문지
2016 .09
-
YOLO v2 Tiny에서 하드웨어 가속화를 위한 Convolution 구조
대한전자공학회 학술대회
2021 .06
-
1-D 수평 Convolution Filter를 이용한 Super Resolution용 Streaming 구조 CNN 하드웨어 최적화
대한전자공학회 학술대회
2018 .06
-
Content-Adaptive Model Update of Convolutional Neural Networks for Super-Resolution
한국방송미디어공학회 학술발표대회 논문집
2020 .11
-
CNN(Convolution Neural Network) 기반의 심혈관질환 분류
정보 및 제어 논문집
2018 .10
-
다중 채널 입력 Convolution residual neural networks 기반의 초해상화 기법
한국방송미디어공학회 학술발표대회 논문집
2016 .06
-
Explainable Grouped Convolution Neural Network Inspired Acoustic-based Mine Recognition for Underwater Surveillance System
한국통신학회 학술대회논문집
2022 .06
-
Convolution Neural Network를 이용한 도로인식 모델
한국정보통신학회 종합학술대회 논문집
2019 .10
-
An Efficient Convolutional Neural Network for Text Image Super-resolution
한국통신학회 학술대회논문집
2018 .01
-
Human Action Recognition Based on 3D Convolutional Neural Network from Hybrid Feature
멀티미디어학회논문지
2019 .12
-
합성곱 신경망 연산을 위한 저전력 콘볼루션 레이어 하드웨어 설계
한국정보통신학회논문지
2024 .07
-
비순차 데이터 입력에 대한 Convolution Neural Network 분류기 적합성 평가
대한전기학회 학술대회 논문집
2022 .07
-
Convolution Neural Network 연구 : Self-Adaptive Voxelization
한국자동차공학회 추계학술대회 및 전시회
2024 .11
-
Deep Convolution Neural Networks를 이용한 PM2.5 농도 추정
대한전자공학회 학술대회
2017 .06
-
Vehicle Detection in Aerial Images using Fully Convolution Regression Neural Network
정보 및 제어 논문집
2018 .10
-
사람 행동 분류를 위한 주의지도 기반 3차원 컨볼루션
제어로봇시스템학회 국내학술대회 논문집
2020 .07
이 논문의 저자 정보
이 논문과 함께 이용한 논문
최근 본 자료
댓글(0)
0
UCI(KEPA) : I410-151-24-02-088527695