Glory Rongyu CHEN

School of Computing, National University of Singapore. rchen@u.nus.edu, glorychen14@gmail.com.

prof_pic.jpg

AS6 #04-11, 11 Computing Drive

Singapore, 117416

Rongyu Chen is a Ph.D. candidate of Computer Science at the Computer Vision & Machine Learning (CVML) Group@National University of Singapore, School of Computing (NUS SoC), fortunately advised by Dr. Angela Yao. He is also mentored by Dr. Linlin Yang. His research interests include,

  • Probabilistic Modeling
  • Generative Models
  • Pose & Shape Estimation applications
  • 3D Reconstruction & Rendering.

He is open to any discussions and collaboration:)

News

Jun, 2024 HANDS@ECCV 2024 is calling for papers & challenge participation!
May, 2024 One co-first author paper “CCNet” was accepted by ICML 2024, VIE, AT, Jul
Jan, 2024 One second author paper “ScaleISH” was accepted by ICLR 2024, VIE, AT, May. Congrats to Kai
Jul, 2023 One paper “MHEntropy” was accepted by ICCV 2023, PAR, FR, Oct
Dec, 2021 One work “Towards Understanding In-Distribution and Out-Of-Distribution of Deep Learning with Deep Generative Models” passed the Qualification Examination

Publications

2024

  1. ccnet_teaser.png
    CCNet: On the Calibration of Human Pose Estimation
    Kerui Gu* ,  Rongyu Chen* ,  (equal) Xuanlong Yu , and  Angela Yao
    In The International Conference on Machine Learning (ICML), 2024
    2D HPE; Uncertainty. A simple yet effective general post-hoc pose confidence estimation calibrates the overlooked original heuristic estimate to improve the mAP & downstream tasks.
  2. scaleish_teaser.png
    ScaleISH: Scaling for Training-Time and Post-Hoc Out-Of-Distribution Detection Enhancement
    Kai Xu ,  Rongyu Chen ,  Gianni Franchi , and  Angela Yao
    In The International Conference on Learning Representations (ICLR), 2024
    OOD Det. The activation scale is a good OOD indicator; neglected post-hoc scaling is better in detecting OOD than commonly used pruning; the scale as "ID-ness" weights training to further improve.

2023

  1. mhentropy_teaser.png
    MHEntropy: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery
    Rongyu Chen ,  Linlin Yang* , and  Angela Yao
    In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023
    3D HMR; Generation. A probabilistic framework deriving the missing entropy and modeling 1-to-m ambiguities with only visible 2D KPs to predict feasible, accurate, & diverse 3D poses.

Activities

  • Reviewed ECCV, ICML, CVPR, AAAI, ICCV, NeurIPS, CVPR, ICLR, BMVC
  • Hosted HANDS@ICCV 2023
  • Attended CVPR 2024 (SEA, US, Jun), ICCV 2023 (PAR, FR, Oct) as a student volunteer, CVPR 2023 (VAN, CA, Jun) with a DEI travel grant, CVPR 2022 (NOLA, US, Jun)

Awards

Oct, 2019 The National Scholarship * 3
Dec, 2018 The ACM-ICPC Asian-East Continent Final Contest, Silver Medal
Dec, 2017 The National College Student Mathematics Competition (CMC) in Provinces, First Prize
Nov, 2011 The National Olympiad in Informatics in Zhejiang Province (NOIP), First Prize