Iasonas Kokkinos
Also known as Iason Kokkinos
Iasonas Kokkinos is a computer-vision researcher whose work spans semantic image segmentation, dense human pose estimation and 3D reconstruction. After earning engineering and doctoral degrees at NTUA and serving on the faculty of École Centrale Paris, he worked as a Research Scientist at Facebook AI Research and coauthored DeepLab, UberNet and DensePose. In 2018, he left FAIR to co-found the augmented reality startup Ariel AI as Chief Executive, and after Snap acquired the company in 2021, he led Snap's Human3D team and coauthored To The Point and MeshPose. A recipient of a 2026 ICBS Frontiers of Science Award for DeepLab, he is an Associate Professor in Computing at Imperial College London, where he researches human-centric physical intelligence.
- Now at
- Imperial College London
Timeline
-
2001–2016
NTUA, UCLA and École Centrale
He earns a D.Eng. at NTUA in 2001, completes a PhD there in 2006, and works as a postdoc at UCLA until 2008. He then joins the faculty of École Centrale Paris, where he stays until 2016, and obtains a Habilitation (HDR) from University Paris-Est in 2013.
-
2016–2018
DeepLab and Facebook AI Research
In 2016, he coauthors DeepLab, which uses atrous convolution and a fully connected conditional random field for semantic image segmentation. From 2016 to 2018, he works as a Research Scientist at Facebook AI Research (FAIR).
-
2017–2018
UberNet and DensePose
At CVPR 2017, he publishes UberNet, an end-to-end convolutional neural network that jointly handles low-, mid- and high-level vision tasks in a unified architecture. At CVPR 2018, he coauthors DensePose, which establishes dense correspondences between RGB-image pixels and a surface-based representation of the human body for dense human pose estimation.
-
2018–2021
Co-founding Ariel AI
In 2018, he leaves FAIR to co-found the London startup Ariel AI as Chief Executive alongside CTO George Papandreou, focusing on real-time monocular human reconstruction for augmented reality. Snap acquires Ariel AI in 2021 to expand Snapchat's augmented reality features.
-
2020–2024
Leading Snap's Human3D team
From 2020 to 2024, he leads the Human3D team at Snap, helping deliver full-body augmented reality experiences to hundreds of millions of Snapchat users.
-
2021–2024
To The Point and MeshPose
In 2021, he coauthors To The Point with Filippos Kokkinos, a method for reconstructing 3D objects from a single image with weakly supervised 2D-to-3D correspondences. At CVPR 2024, he coauthors MeshPose, which uses weak DensePose supervision to localize mesh vertices in 2D and lift them into a low-poly 3D body mesh for real-time AR.
-
2026
ICBS Frontiers of Science Award
In 2026, Kokkinos and his DeepLab coauthors receive the Frontiers of Science Award from the International Congress for Basic Science.
-
Since 2026
Associate Professor at Imperial
After serving as an Associate Professor at University College London, he joins Imperial College London in September 2026 as an Associate Professor in Computing in the Department of Computing. At Imperial, he researches human-centric physical intelligence and supervises work in human-centric robot learning.
Career
- NowImperial College LondonAssociate Professor in Computing
- Until 2026University College LondonAssociate Professor
- 2020–2024SnapTeam Lead
- 2018–2021Ariel AICEO, Chief Executive
- Until 2016École Centrale ParisFaculty Member
- Until 2008University of California, Los AngelesPostdoctoral Researcher
- FAIR (Meta)Research Scientist
Education
- 2013University Paris-EstHabilitation
- 2006National Technical University of AthensPhD
- 2001National Technical University of AthensD.Eng.
Sources and credits
Photo: We have no published photo of Iasonas Kokkinos yet. Send us a free-licensed photo.
- [Archimedes Talks Series] Learning-based 3D understanding of deformable objects using weak and self-supervision
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs
- Iasonas Kokkinos
- UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory
- DensePose: Dense Human Pose Estimation in the Wild
- Snap Buys UK Tech Startup Ariel AI, Signaling Push Into AR
- Snap acquires Ariel AI to boost Snapchat augmented reality features
- To The Point: Correspondence-driven monocular 3D category reconstruction
- MeshPose: Unifying DensePose and 3D Body Mesh Reconstruction
- Yuille Team Receives 2026 ICBS Frontiers of Science Award (CCVL, Johns Hopkins)
- Iasonas Kokkinos | alphaXiv
- People | Faculty of Engineering | Imperial College London
- Scholarships | Faculty of Engineering | Imperial College London

