Youwei Pang 庞有伟
Research Fellow · CCDS, Nanyang Technological University
My long-term goal is visual intelligence that generalizes across scales, tasks and modalities, stays efficient from large models down to real deployments, and earns trust through reliable, diagnostic evaluation — so that what a model perceives can be measured, compared and relied on.
01
Generalizable
- Multi-scale perception — mixed-scale, high-res, tiny targets
- Unified multi-task — context-dependent concepts
- Multimodal — RGB-D/T, video, open-vocabulary, missing modality
02
Efficient
- Architectures — efficient attention, dynamic operators
- Deployment — large model feature coding, lithium battery & semiconductor inspection
03
Reliable
- Evaluation — metric decomposition, diagnosis, toolkits
- Benchmarks — X-ray inspection, feature coding, agents
- Content safety — harmful content
Off the clock: coding, reading, music, running, swimming, city walks and good food. Most projects are built together with my best friend Xiaoqi Zhao.
Trajectory
-
2015 – 2019B.E. Electronic Information Engineering
-
2019 – 2025Ph.D. Signal & Information Processing, with Prof. Lihe Zhang & Prof. Huchuan Lu
-
2025.09 – nowResearch Fellow, CCDS, with Prof. Weisi LinLarge model feature coding · Harmful content detection
Code
GitHubHonors
- 2025–26World's Top 2% Scientists (Stanford / Elsevier)
- 2026Outstanding Reviewer, ACL 2026
- 2023Gold Award, 8th China International College Students' "Internet+" Innovation & Entrepreneurship Competition (161 / 3.4M+)
- 2022MICCAI GOALS Challenge 2 / 100 — Glaucoma Detection 1st, OCT Layer Segmentation 3rd
News
Resources
Service
Program Committee / Reviewer
- CVPR
- 22–26
- ICCV
- 21–25
- ECCV
- 22–26
- NeurIPS
- 23–26
- ICLR
- 24–27
- ICML
- 25
- ACM MM
- 25–26
- AAAI
- 22, 27
- ACL
- 26
- WACV
- 27
- BMVC
- 26
- ACCV
- 26
Journal Reviewer
Guest Editor
Journal of Imaging Special Issue: Unified and Multimodal Segmentation: Foundations, Methods, and Applications Submissions welcome
Talks
VALSE 2024, Chongqing