Lin GU 谷林
Research Scientist, RIKEN AIP
Special Researcher, the University of Tokyo
I am now a research scientist at RIKEN AIP, Japan (理化学研究所), as well as a special researcher at the University of Tokyo (東京大学). Prior to my current positions, I joined the National Institute of Informatics (国立情報学研究所) in Tokyo in June 2016. I also served as a regular visiting scholar at Kyoto University (京都大学 ) from 2016 to 2019. Before moving to Japan, I held the position of Postdoctoral Research Fellow at the Bioinformatics Institute, A*STAR, Singapore. My doctoral studies were completed at the Australian National University and NICTA (Now Data61) in 2014. During which I focused on hyperspectral imaging and colour science.
I have authored over 60 papers in journals or conferences such as Nature Methods, IJCV, AAAI, CVPR, ECCV, ICCV and MICCAI. I am an associate editor for Pattern Recognition and area chair for ICML 24, Neruips 24.
Research Interest:
My research spans a diverse array of subjects within the realm of computer vision and deep learning. It is driven by two primary motivations: advancing core artificial intelligence research and leveraging its applications across diverse disciplines. The initial impetus revolves around trailblazing machine learning algorithms capable of learning effectively from limited and less-than-ideal data. Building upon fundamental machine learning, I collaborate with specialists across multiple domains to catalyse transformative research endeavours. These collaborative efforts extend across fields, including the medical domain, nuclear fusion, and Terahertz imaging.
Ongoing Research
Presently, I hold the role of a project manager（課題推進者） within the Moonshot Program Moonshot Program (ムーンショット型研究開発制度 ). My primary focus revolves around spearheading research related to Continuous Learning and Memory Mechanisms (継続学習と記憶の研究開発). This initiative exploits insights from diverse disciplines , from neuroscience to physics, to endow existing artificial intelligence (AI) with memory ability. The final goal is to equip them with the capability to effectively navigate through intricate and non-ideal scenarios in the real world.
Moonshot is a national program overseen by the Cabinet of Japan (内閣府) that seeks to stimulate high-risk, high-impact research and development, ultimately achieving innovative value creation tailored to Japan's needs in the modern era.The program is designed to advance ambitious R&D projects with the explicit purpose of addressing intricate societal challenges. Notably, I am engaged in Goal 1, which aims to manifest a society where, by 2050, individuals are liberated from the constraints of body, brain, space, and time (2050年までに、人が身体、脳、空間、時間の制約から解放された社会を実現).
I currently serve as the Principal Investigator (PI) representing Japan for the RIKEN-MOST project, which centres on subtyping schizophrenia through artificial intelligence techniques. This $600,000 project harnesses state-of-the-art AI technology to delve into various domains such as gaze analysis, EEG signals, facial gestures, and fMRI data. Moreover, it integrates diverse omics data, spanning genome, proteome, transcriptome, epigenome, and metabolome data, to comprehensively investigate schizophrenia. The ultimate goal is to establish a solid groundwork for both the early diagnosis and subtyping of schizophrenia and other mental disorders.
Recruiting researchers and research assistants at RIKEN AIP for my Moonshot projects.
https://www.riken.jp/careers/researchers/20240614_2/index.html
News:
I am honoured to serve as an Area Chair for Neurips 2024
I've joined the Pattern Recognition Journal as an Associate Editor.
Our paper Exploring the Usage of Pre-trained Features for Stereo Matching has been accepted by IJCV.
Our paper Adaptive Dilated Convolution from Frequency View has been selected as a highlight (2.8%) at this year's CVPR.
Our paper Content-Specific Humorous Image Captioning Using Incongruity Resolution Chain-of-Thought has been accepted at NAACL 2024.
We have three papers accepted at CVPR 2024.
I am honoured to serve as Area Chair for ICML 2024
We have one paper accepted at ICLR 2024.
When Semantic Segmentation Meets Frequency Aliasing
We have one paper accepted at AAAI 2024.
Aleth-NeRF: Illumination Adaptive NeRF with Concealing Field Assumption
Our journal paper, Sketch-based semantic retrieval of medical images, is published on Medical Image Analysis
We have two papers (one oral and one poster) accepted for ICCV 2023.
Our journal paper, BigNeuron: a resource to benchmark and predict the performance of algorithms for automated tracing of neurons in light microscopy datasets, is published by Nature Methods.
RIKEN has issued a news release about our ECCV 2022 paper: Exploring Resolution and Degradation Clues as Self-supervised Signal for Low Quality Object Detection.