An Efficient Frequency Domain Vision Pipeline From RAW Images to Backend Tasks

doi:10.1109/ISCAS46773.2023.10182018

	An Efficient Frequency Domain Vision Pipeline From RAW Images to Backend Tasks
	Haoyan Li; Wei Zhou; Xiangyu Zhang; Xin Lou
	2023
会议录名称	PROCEEDINGS - IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS
ISSN	0271-4310
卷号	2023-May
发表状态	已发表
DOI	10.1109/ISCAS46773.2023.10182018
摘要	Though high resolution benefits computer vision performance, they are not commonly used in convolutional neural network (CNN)-based vision algorithms due to the limitation of memory and computation resource. Learning in the frequency domain makes high resolution images directly acceptable by CNNs, but the computation, time and energy overhead for pre-processing, including image signal processing (ISP) and domain transformation, can be large. This paper explores different image processing and domain transformation operations and proposes an efficient end-to-end frequency domain learning pipeline from RAW images to vision tasks. In particular, we simplify the pre-processing part by skipping the entire ISP pipeline and replacing the Discrete Cosine Transform (DCT) with a multiplication-free approximated one. Experimental results show that the final vision performance of the proposed pipeline is very close to that of the conventional pipeline, while significant amount of redundant operations can be saved. © 2023 IEEE.
会议举办国	et al.; Huawei; IEEE; IEEE Circuits and Systems Society (CAS); Samsung Semiconductor; Synopsys
会议录编者/会议主办者	et al. ; Huawei ; IEEE ; IEEE Circuits and Systems Society (CAS) ; Samsung Semiconductor ; Synopsys
关键词	Computer vision Convolutional neural networks Frequency domain analysis Image compression Pipeline processing systems Pipelines Discrete cosine transform Domain learning Domain transformation Frequency domain learning Frequency domains High resolution Image signal processing Performance Pre-processing
会议名称	56th IEEE International Symposium on Circuits and Systems, ISCAS 2023
会议地点	Monterey, CA, United states
会议日期	May 21, 2023 - May 25, 2023
URL	查看原文
收录类别	EI
语种	英语
出版者	Institute of Electrical and Electronics Engineers Inc.
EI入藏号	20233314552700
EI主题词	Discrete cosine transforms
EI分类号	619.1 Pipe, Piping and Pipelines ; 722.4 Digital Computers and Systems ; 723.5 Computer Applications ; 741.2 Vision ; 921.3 Mathematical Transformations
原始文献类型	Conference article (CA)
来源库	IEEE
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/325827
专题	信息科学与技术学院信息科学与技术学院_PI研究组_娄鑫组信息科学与技术学院_博士生
作者单位	School of Information Science and Technology, ShanghaiTech University, Shanghai, China
第一作者单位	信息科学与技术学院
第一作者的第一单位	信息科学与技术学院
推荐引用方式 GB/T 7714	Haoyan Li,Wei Zhou,Xiangyu Zhang,et al. An Efficient Frequency Domain Vision Pipeline From RAW Images to Backend Tasks[C]//et al., Huawei, IEEE, IEEE Circuits and Systems Society (CAS), Samsung Semiconductor, Synopsys:Institute of Electrical and Electronics Engineers Inc.,2023.