Haodong Duan 段浩东
Haodong Duan is a postdoctoral researcher at Shanghai AI Lab, working on the evaluation of large language models and multi-modality models. He received his Ph.D. degree from the Multimedia Laboratory @ CUHK in 2023, supervised by Professor Dahua Lin. During his Ph.D., he works on human-centric action understanding, including video classification, skeleton-based action recognition, pose estimation. Before joining MMLAB, he received his B.S. degree in data science at Peking University in 2019. His research interests include video recognition, human-centric action understanding, and multi-modality learning. You can find his CV here.
News
- (2024.04) We released MMStar
. In this work, we studied the vision-dispensable and data leakage problems in multi-modal evaluation
- (2024.03) Two papers (BotChat
, Ada-LEval
) are accepted by NAACL 2024
- (2023.12) We released VLMEvalKit
, an all-in-one toolkit for evaluating LVLMs
- (2023.10) SkeleTR is accepted by ICCV 2023
- (2023.07) We released MMBench, a pioneering benchmark for evaluating large vision-language models (LVLMs)
- (2023.06) I received my Ph.D. degree from the Multimedia Laboratory @ CUHK
- (2023.04) I will join the Shanghai AI Lab as a postdoctoral researcher
- (2022.07) I started my internship at AWS AI, advised by Dr. Mingze Xu
- (2022.06) Give a talk at CVPR 2022 OpenMMLab Tutorial on human-centric action understanding [Slides]
- (2022.05) Release PYSKL
, a codebase for skeleton action recognition [Report] accepted by MM 2022
- (2022.03) 3 papers (PoseC3D, TransRank, OCSampler) are accepted by CVPR 2022
- (2020.08) Join OpenMMLab and serve as a maintainer of MMAction2
- (2020.07) OmniSource is accepted by ECCV 2020
- (2019.07) TRB is accepted by ICCV 2019
Bold indicates being accepted as Oral presentation.
Professional Activities
- Conference Reviewer: ICCV[21-23], AAAI[22-24], CVPR[22-24], ECCV[22-24], NeurIPS[22-23], WACV23, ICLR23, EuroGraphics, etc.
- Journal Reviewer: TPAMI, IJCV, TIP, PR, TMM, etc.