Ross Girshick
First author of R-CNN and Detectron; co-author of Segment Anything; formerly at FAIR, Microsoft and Ai2
Ross Girshick's research interests include algorithms for visual perception, representation learning, and the interaction of vision and language. Girshick's site lists research scientist roles at Microsoft Research (2014–2015), FAIR at Facebook/Meta (2015–2023), and the Allen Institute for AI (2023–2024), and notes that Girshick co-founded Vercept in late 2024, a company acquired in early 2026. Girshick is first author of the 2013 R-CNN object detection paper, is listed first in the citation entry for Detectron, FAIR's Apache-2.0 detection system, and appears among the authors of its successor Detectron2. Girshick is also a co-author of the original Segment Anything paper and of SAM 2.126734
Work
Segment Anything (SAM and SAM 2)Open weights
Co-author of the Segment Anything paper (2023) and the SAM 2 paper (2024). The original repository lists Girshick among its authors and releases the model under the Apache 2.0 license.345
Detectron and Detectron2 (external site: github.com)Open-source code
Listed first in the citation entry of Detectron, FAIR's object detection system written in Python on Caffe2 and released under Apache 2.0, and listed among the authors in the citation entry of the Apache-2.0 Detectron2 library.67
R-CNN (external site: arxiv.org)Published research
First author, with Jeff Donahue, Trevor Darrell, and Jitendra Malik, of "Rich feature hierarchies for accurate object detection and semantic segmentation" (2013), the R-CNN paper.2
Other documented roles
Roles the sources document but that are not confirmed as current.
- Research scientist, FAIR (2015–2023), Meta (Facebook AI Research)1
- Research scientist (2023–2024), Allen Institute for AI1
- Research scientist (2014–2015), Microsoft Research1
- Co-founder (late 2024; the company was acquired in early 2026), Vercept1
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.