Tsung-Yi Lin
Principal research scientist at NVIDIA Research; led the creation of the COCO dataset; Cosmos co-author
Tsung-Yi Lin is a principal research scientist at NVIDIA Research, working on computer vision, machine learning, and generative AI, with a recent interest in generative AI in 3D; Lin was previously on the Brain Team at Google Research. Lin's site says Lin led the creation of the COCO (Common Objects in Context) dataset, and Lin is first author of the 2014 Microsoft COCO paper and of the 2017 Focal Loss paper. Lin is also among the authors of NVIDIA's 2025 Cosmos World Foundation Model Platform paper, which describes Cosmos as open-source with open-weight models.1345
Work
NVIDIA CosmosOpen weights
Listed among the authors of "Cosmos World Foundation Model Platform for Physical AI" (2025), which says the platform is open-source and its models are open-weight under permissive licenses. Lin's NVIDIA Research profile page also lists Cosmos-Reason 1 (2025) among Lin's publications.52
COCO dataset (external site: cocodataset.org)Open data
First author of "Microsoft COCO: Common Objects in Context" (2014); Lin's site says Lin led the creation of the dataset.31
Focal Loss (external site: arxiv.org)Published research
First author of "Focal Loss for Dense Object Detection" (2017) with Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár.4
Other documented roles
Roles the sources document but that are not confirmed as current.
- Member of the Brain Team, Google Research, Google Research (Brain Team)1
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.