Piotr Dollár

Career

Piotr Dollár completed a PhD in computer science at the University of California, San Diego in 2007 with a dissertation on learning from local image regions, then held a postdoctoral fellowship at Caltech until 2011. From 2010 to 2011 Dollár was a cofounder of Anchovi Labs, which Dropbox acquired in 2012.⁹¹

Dollár was a researcher at Microsoft Research from November 2011 to September 2014 and has been a research director at Facebook AI Research since September 2014. Meta's profile lists research interests in recognition, grouping, network design, and self-supervised learning. Dollár has served as an area chair for CVPR, ICCV, ECCV, and NeurIPS.⁹²

Research

Dollár co-authored the 2014 paper introducing Microsoft COCO, a dataset for object recognition in everyday scenes, and in 2017 co-authored Feature Pyramid Networks, Focal Loss for dense object detection, and Mask R-CNN. Mask R-CNN and related detection work were released in FAIR's Detectron codebase, whose citation entry lists Dollár among the authors.⁶⁷⁹⁸

Later work includes Masked Autoencoders Are Scalable Vision Learners (CVPR 2022) and the Segment Anything line of models. Dollár co-authored Segment Anything in 2023 and SAM 2 in 2024, is an author of the SAM 3 paper from November 2025, and is named among the project leads in the SAM 3 repository.⁹³⁴¹⁰⁵

Work and contributions

• Segment Anything (SAM, SAM 2 and SAM 3) Open weights

  Co-author of the Segment Anything paper (2023) and the SAM 2 paper (2024), an author of the SAM 3 paper (2025), and one of the project leads named in the SAM 3 repository README.³⁴⁵¹⁰

• COCO dataset Open data

  Co-author of "Microsoft COCO: Common Objects in Context" (2014), the paper that introduced the dataset.⁶

• Mask R-CNN and Detectron Open-source code

  Co-author of the Mask R-CNN paper (2017) and listed among the authors in the citation entry of Detectron, FAIR's Apache-2.0 object detection system.⁷⁸

• Piotr's Image and Video Matlab Toolbox Open-source code

  Author of an open-source computer vision library for Matlab with an emphasis on recognition, alongside other toolboxes for behavior recognition, cascaded pose regression, and structured edge detection.⁹

Timeline

1. 2007Completed a PhD in computer science at UC San Diego on learning from local image regions.⁹
2. Jul 2010Cofounded Anchovi Labs, later acquired by Dropbox.⁹
3. Nov 2011Joined Microsoft Research as a researcher.⁹
4. May 1, 2014Co-authored Microsoft COCO, Common Objects in Context.⁶
5. Sep 2014Became a research director at Facebook AI Research.⁹
6. Mar 20, 2017Co-authored the Mask R-CNN paper, later presented at ICCV 2017.⁷⁹
7. Apr 5, 2023Co-authored Segment Anything.³
8. Aug 1, 2024Co-authored SAM 2, Segment Anything in Images and Videos.⁴
9. Nov 20, 2025Co-authored the SAM 3 paper and is listed as a project lead in the SAM 3 repository.¹⁰⁵

Other documented roles

Roles the sources document that are not confirmed as current.

• Research at Microsoft Research, before joining FAIR in 2014, Microsoft Research¹
• Helped co-found the company (acquired by Dropbox in 2012), Anchovi Labs¹
• Postdoctoral Fellow (2007–2011), California Institute of Technology⁹

Honors and recognition

• ICCV Best Paper Award (Marr Prize) for Mask R-CNN · International Conference on Computer Vision (ICCV), 2017⁹
• ICCV Best Student Paper Award for Focal Loss for Dense Object Detection · International Conference on Computer Vision (ICCV), 2017⁹
• CVPR Best Paper Honorable Mention for Segment Anything · Conference on Computer Vision and Pattern Recognition (CVPR), 2023⁹
• PAMI Mark Everingham Prize for Microsoft COCO · IEEE PAMI Technical Committee⁹

Where they work

• Meta 106 people in the wiki

Related people

• Matthijs DouzeAlso at Meta
• Matt DeitkeAlso at Meta
• Sachin MehtaAlso at Meta
• Yuxin WuAlso at Meta
• Adina WilliamsAlso at Meta
• Tsung-Yi LinComputer vision · Datasets and data
• Alexander ToshevComputer vision · Datasets and data
• Fei-Fei LiComputer vision · Datasets and data
• Geoffrey HintonComputer vision
• Shusheng YangComputer vision