Martin Danelljan
Martin Danelljan is a computer vision researcher whose work on visual object tracking and 3D vision runs from Linköping University through ETH Zurich to Apple. With Michael Felsberg, Fahad Khan and Gustav Häger, he won the 2014 Visual Object Tracking Challenge, defended his Linköping doctoral thesis in 2018, and received the Best Nordic Thesis Prize at SCIA 2019. He coauthored the ATOM and discriminative model prediction trackers in 2019 and the RePaint diffusion inpainting method in 2022, and at ETH Zurich he was a postdoc and then a research group leader. Since November 2023, he is a senior computer vision research engineer in Apple's 3D Vision group, where he builds a team working on 3D computer vision at a global scale.
Timeline
-
2014
Winning the VOT Challenge
A prize-winning degree project in object tracking leads him to doctoral studies at Linköping University. In 2014, he wins the Visual Object Tracking Challenge with Michael Felsberg, Fahad Khan and Gustav Häger, and also the OpenCV State-of-the-Art Vision Challenge.
-
2016
Best paper at ICPR
He receives the best paper award at ICPR 2016 and achieves a top rank in the VOT2016 challenge.
-
2018
Doctorate and Singulareye
In June 2018, he defends his thesis, Learning Convolution Operators for Visual Tracking, as the first doctoral thesis associated with the WASP research programme, after working at Linköping's Computer Vision Laboratory. In 2018, he also founds the company Singulareye with two other doctoral students.
-
2019
The ATOM and DiMP trackers
In 2019, he is first author of ATOM, a CVPR 2019 tracker that predicts the overlap between a target and its bounding box and gains 15% over the previous best approach on TrackingNet at over 30 FPS. He also coauthors Learning Discriminative Model Prediction for Tracking with Goutam Bhat, Luc Van Gool and Radu Timofte, an end-to-end tracker that sets a new state of the art on 6 benchmarks.
-
2019
Two awards in 2019
His doctoral thesis wins the Best Nordic Thesis Prize at SCIA 2019. He also receives the best student paper award at BMVC 2019 for Tracking the Known and the Unknown by Leveraging Semantic Information, written with Ardhendu Shekhar Tripathi, Luc Van Gool and Radu Timofte.
-
2019–2023
Group leader at ETH Zurich
From 2019 to 2021, he is a postdoc at ETH Zurich in Radu Timofte's team, with Luc Van Gool as co-supervisor, and his research covers online and meta-learning for visual tracking and video object segmentation and deep probabilistic models for image generation. He goes on to lead a research group at ETH Zurich and founds Elevate3D.
-
2022
The RePaint inpainting method
In 2022, he coauthors RePaint with Andreas Lugmayr, Andres Romero, Fisher Yu, Radu Timofte and Luc Van Gool. The method is a denoising diffusion probabilistic model approach to inpainting that applies to even extreme masks.
-
Since November 2023
Research engineer at Apple
He joins Apple in November 2023 as a senior computer vision research engineer in the 3D Vision group. There he builds a team working on 3D computer vision at a global scale.
Career
- Since 2023AppleSenior Computer Vision Research Engineer
- 2019–2023ETH ZurichPostdoc, Research Group Leader
- Until 2018Linköping University
- From 2018SingulareyeCo-founder
- Elevate3DFounder
Education
Sources and credits
Photo: We have no published photo of Martin Danelljan yet. Send us a free-licensed photo.
- Highly cited LiU articles on computer vision - Linköping University
- From a degree project to 9000 citations and a company | ELLIIT
- Visual tracking: from correlation filters to end-to-end discriminative model prediction
- Martin Danelljan defends first WASP associated doctoral thesis | WASP
- LiU a leader within AI - Linköping University
- [1811.07628] ATOM: Accurate Tracking by Overlap Maximization
- [1904.07220] Learning Discriminative Model Prediction for Tracking
- Radu Timofte, Computer Vision Lab, University of Wurzburg
- Martin Danelljan
- [2201.09865] RePaint: Inpainting using Denoising Diffusion Probabilistic Models



