Alexander Kirillov

Career

Kirillov's homepage records doctoral work at Heidelberg University under the supervision of Carsten Rother, after a diploma at Lomonosov Moscow State University with Dmitry Vetrov and Alexander Dyakonov. Kirillov then became a research scientist at Facebook AI Research (FAIR), later led the multimodal research group at OpenAI that worked on Advanced Voice Mode and GPT-4o, and now works on multimodal AI models at Thinking Machines.¹

Research

Much of Kirillov's research concerns image segmentation. "Panoptic Segmentation" (2018), with Kaiming He, Ross Girshick, Carsten Rother and Piotr Dollár, proposed a task that unifies semantic and instance segmentation along with a panoptic quality metric. Follow-up work included Panoptic Feature Pyramid Networks and PointRend, which treats segmentation as a rendering problem.⁶⁷⁸

Kirillov co-authored DETR, which frames object detection as direct set prediction with a transformer, and later work on universal image segmentation with a masked-attention mask transformer. As first author of Segment Anything (2023), Kirillov introduced a promptable segmentation model released with a large segmentation dataset and its code.⁵¹²³

Work and contributions

• Segment Anything (SAM) Open weights

  First author of the Segment Anything paper (2023); the original repository lists Kirillov first among its authors and releases the model under the Apache 2.0 license.²³

• Detectron2 Open-source code

  Listed among the authors in the citation entry of Detectron2, FAIR's detection and segmentation library released under the Apache 2.0 license.⁴

• DETR Open-source code

  Co-author of "End-to-End Object Detection with Transformers" (2020); the DETR repository is released under the Apache 2.0 license.⁵⁹

Timeline

1. 2015Presented "Inferring M-Best Diverse Labelings in a Single One" at ICCV 2015 as first author.¹
2. Jan 2018Posted "Panoptic Segmentation" as first author with Kaiming He, Ross Girshick, Carsten Rother and Piotr Dollár.⁶
3. Jan 2019Posted "Panoptic Feature Pyramid Networks" as first author.⁷
4. Dec 2019Posted "PointRend: Image Segmentation as Rendering" as first author.⁸
5. May 2020Co-authored "End-to-End Object Detection with Transformers" (DETR).⁵
6. 2022Co-authored "Masked-attention Mask Transformer for Universal Image Segmentation" at CVPR 2022.¹
7. Apr 2023Posted the Segment Anything paper as first author, with the model released under Apache 2.0.²³

Other documented roles

Roles the sources document that are not confirmed as current.

• Led the multimodal research group (Advanced Voice Mode and GPT-4o), OpenAI¹
• Research scientist, FAIR, Meta (Facebook AI Research)¹

Where they work

• Thinking Machines Lab 13 people in the wiki

Related people

• Mira MuratiAlso at Thinking Machines Lab
• Danqi ChenAlso at Thinking Machines Lab
• John SchulmanAlso at Thinking Machines Lab
• Soumith ChintalaAlso at Thinking Machines Lab
• Horace HeAlso at Thinking Machines Lab
• Jyoti AnejaMultimodal models · Computer vision
• Cordelia SchmidComputer vision · Multimodal models
• Ali FarhadiComputer vision · Multimodal models
• Christopher ClarkMultimodal models · Computer vision
• Fei-Fei LiComputer vision · Multimodal models