Learned Smartphone ISP on Mobile GPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report

doi:10.1007/978-3-031-25066-8_3

	Learned Smartphone ISP on Mobile GPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report
	Ignatov, Andrey 1,2; Timofte, Radu 1,2,3; Liu, Shuai 4; Feng, Chaoyu 4; Bai, Furui 4; Wang, Xiaotao 4; Lei, Lei 4; Yi, Ziyao 5; Xiang, Yan 5; Liu, Zibin 5; Li, Shaoqing 5; Shi, Keming 5; Kong, Dehui 5; Xu, Ke 5; Kwon, Minsu 6; Wu, Yaqi 7; Zheng, Jiesi 8; Fan, Zhihao 9; Wu, Xun 10; Zhang, Feng 7,8,9,10; No, Albert 11; Cho, Minhyeok 11; Chen, Zewen 12; Zhang, Xiaze 13; Li, Ran 14; Wang, Juan 12; Wang, Zhiming 10; Conde, Marcos V.3; Choi, Ui-Jin 3; Perevozchikov, Georgy 15; Ershov, Egor 15; Hui, Zheng 16; Dong, Mengchuan17 ; Lou, Xin 1; Zhou, Wei17 ; Pang, Cong17 ; Qin, Haina 12; Cai, Mingxuan 12
	2023
会议录名称	LECTURE NOTES IN COMPUTER SCIENCE (INCLUDING SUBSERIES LECTURE NOTES IN ARTIFICIAL INTELLIGENCE AND LECTURE NOTES IN BIOINFORMATICS)
ISSN	0302-9743
卷号	13803 LNCS
页码	44-70
发表状态	已发表
DOI	10.1007/978-3-031-25066-8_3
摘要	The role of mobile cameras increased dramatically over the past few years, leading to more and more research in automatic image quality enhancement and RAW photo processing. In this Mobile AI challenge, the target was to develop an efficient end-to-end AI-based image signal processing (ISP) pipeline replacing the standard mobile ISPs that can run on modern smartphone GPUs using TensorFlow Lite. The participants were provided with a large-scale Fujifilm UltraISP dataset consisting of thousands of paired photos captured with a normal mobile camera sensor and a professional 102MP medium-format FujiFilm GFX100 camera. The runtime of the resulting models was evaluated on the Snapdragon’s 8 Gen 1 GPU that provides excellent acceleration results for the majority of common deep learning ops. The proposed solutions are compatible with all recent mobile GPUs, being able to process Full HD photos in less than 20–50 ms while achieving high fidelity results. A detailed description of all models developed in this challenge is provided in this paper. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
关键词	Cameras Deep learning Image enhancement Large dataset Program processors AI benchmark Deep learning Fujifilms Image signal processing Learned image signal processing Mobile AI Mobile AI challenge Mobile camera Photo enhancement Smart phones
会议名称	17th European Conference on Computer Vision, ECCV 2022
会议地点	Tel Aviv, Israel
会议日期	October 23, 2022 - October 27, 2022
收录类别	EI
语种	英语
出版者	Springer Science and Business Media Deutschland GmbH
EI入藏号	20231413826863
EI主题词	Smartphones
EISSN	1611-3349
EI分类号	461.4 Ergonomics and Human Factors Engineering ; 718.1 Telephone Systems and Equipment ; 723.2 Data Processing and Image Processing ; 742.2 Photographic Equipment
原始文献类型	Conference article (CA)
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/292237
专题	信息科学与技术学院_硕士生信息科学与技术学院_博士生
通讯作者	Ignatov, Andrey
作者单位	1.Computer Vision Lab, ETH Zurich, Zurich, Switzerland; 2.AI Witchlabs, Zurich, Switzerland; 3.University of Wuerzburg, Wuerzburg, Germany; 4.Xiaomi Inc., Beijing, China; 5.Sanechips Co. Ltd., Shanghai, China; 6.ENERZAi, Seoul, Korea, Republic of; 7.Harbin Institute of Technology, Harbin, China; 8.Zhejiang University, Hangzhou, China; 9.University of Shanghai for Science and Technology, Shanghai, China; 10.Tsinghua University, Beijing, China; 11.Hongik University, Seoul, Korea, Republic of; 12.Institute of Automation, Chinese Academy of Sciences, Beijing, China; 13.School of Computer Science, Fudan University, Shanghai, China; 14.Washington University in St. Louis, Seattle, United States; 15.Moscow Institute of Physics and Technology, Moscow, Russia; 16.Alibaba DAMO Academy, Beijing, China; 17.ShanghaiTech University, Shanghai, China
推荐引用方式 GB/T 7714	Ignatov, Andrey,Timofte, Radu,Liu, Shuai,et al. Learned Smartphone ISP on Mobile GPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report[C]:Springer Science and Business Media Deutschland GmbH,2023:44-70.