About
Jin Ye is a PhD student at Monash University (Melbourne, Australia), Department of Data Science and AI, supervised by Prof. Jianfei Cai, A/Prof. Zhaolin Chen, and Dr. Bohan Zhuang. Before starting his PhD, he was a full-time researcher at Shanghai AI Laboratory, where he worked as a core member of the GMAI team under Dr. Junjun He and Prof. Yu Qiao. Earlier in his career, he was a Senior R&D Engineer at Baidu, collaborating on computer vision engineering.
His research sits at the intersection of medical image segmentation, medical vision-language models, and generative modeling, with the goal of translating advanced AI into real clinical practice. Within the GMAI team, he was the principal architect behind the SA-Med2D-20M dataset and a key contributor to the SAM-Med series.
Research Interests
- Large-scale medical segmentation datasets: SA-Med2D-20M (4.6M images / 19.7M masks)
- SAM adaptation for medicine: SAM-Med2D, SAM-Med3D, SAM-Med3D-MoE
- Medical AI benchmarking: GMAI-MMBench (284 datasets, 38 modalities, 18 tasks)
- Interactive medical image segmentation: IMed-361M
Selected Publications
- SA-Med2D-20M (arXiv 2023) — First Author. The largest 2D medical segmentation dataset at the time; foundational data resource for the SAM-Med research line.
- SAM-Med2D (arXiv 2023) — Comprehensive study of SAM adaptation to 2D medical images via learnable adapters. GitHub repository has 3,900+ stars.
- SAM-Med3D (ECCV 2024 Workshop Oral; IEEE TNNLS 2025) — Extension of SAM to 3D volumetric medical images using SA-Med3D-140K.
- SAM-Med3D-MoE (MICCAI 2024) — Mixture-of-Experts extension to prevent catastrophic forgetting in task-specific finetuning.
- GMAI-MMBench (NeurIPS 2024) — 2nd Author. The most comprehensive medical AI benchmark to date.
- IMed-361M (CVPR 2025) — Interactive medical segmentation benchmark and baseline.
Career Trajectory
Baidu (Senior R&D Engineer) → Shanghai AI Laboratory (Researcher, GMAI Team) → Monash University (PhD Student) — a path bridging applied engineering, large-scale research infrastructure, and academic depth.