Intro

I'm a Software Architect at Maritime Robotics, working with distributed software systems for Unmanned Surface Vessels. I did my Ph.D. at NTNU working with application of computer vision and generative models for realistic image anonymization. My work in this area has been recognized with several awards, including the best AI master thesis from the Norwegian Open AI Lab, best paper at ISVC2019, and the best demo at NorwAI Innovate 2022.


All my work is published open-source, so check out my github page!.

Education

Ph.D. Jun. 2019 - Jun. 2023
Norwegian University of Science and Technology

Topic: Generative models for image anonymization to ensure privacy and usability of image data for computer vision development.

M.Sc. Computer Science Aug. 2014 - Jun. 2019
Norwegian University of Science and Technology

Thesis title: DeepPrivacy: A GAN-based framework for image anonymization.

University of California, San Diego Aug. 2017 - Jun 2018
Study abroad year

Work Experience

Maritime Robotics November 2024 - Now
Software Architect

Maritime Robotics August 2023 - November 2024
Senior Computer Vision / Machine Learning Engineer

Cisco Systems Summer 2018 & 19
Machine Learning Intern

Summer 2019: Development of real-time, low-footprint (CPU-based) face detection.
Summer 2018: Development of low-footprint keyword spotting detector ("Ok Google"-like).

Itera Summer 2017
IT Consultant Intern

Responsible for intergation to third-party services for managing and distributing digital door keys to phones.

NTNU, Department of Civil and Environmental Engineering Dec. 2016 - Aug. 2017
Research Assistant

Implemented a React-based smart-watch app to ease the reading of critical values connected to swimming halls.

Teaching Experience

NTNU, Department of Computer Science Aug. 2018 - Jun 2022
Head Teaching Assistant

Responsible for mandatory course work and assignment lectures in the following courses:
  TDT4265 - Computer Vision and Deep Learning
  TDT4195 - Visual Computing Fundamentals

NTNU, Department of Computer Science Aug. 2015 - Jun 2017
Teaching Assistant

Support courses by advising cstudents, grading exercises, lecturing and developing assignments in the following courses:
  TDT4145 - Data Modelling, Databases and Database Management Systems
  TDT4120 - Algorithms and Data Structures
  TDT4100 - Object Oriented Programming
  TMA4100 - Calculus 1

Achievements

ISVC 2019 best paper award for the paper: DeepPrivacy: A Generative Adversarial Network for Face Anonymization.

NTNU OpenAI Lab Best AI Master Thesis for the thesis: DeepPrivacy: A GAN-based framework for image anonymization.

NorwAI Innovate 2022 best demo/poster for the demo: DeepPrivacy2: A Framework for Realistic Image Anonymization.

Publications

See my Google Scholar profile for an updated list.

Does Image Anonymization Impact Computer Vision Training? CVPR 2023 WAD
Håkon Hukkelås, Frank Lindseth [PDF] [Github]

Synthesizing Anyone, Anywhere, in Any Given Pose WACV 2024
Håkon Hukkelås, Frank Lindseth [PDF] [Github]

DeepPrivacy2: Towards Realistic Full-Body Anonymization WACV 2023
Håkon Hukkelås, Frank Lindseth [PDF] [Github]

Realistic Full-Body Anonymization with Surface-Guided GANs WACV 2023
Håkon Hukkelås, Morten Smebye, Rudolf Mester, Frank Lindseth [PDF] [Github]

Image inpainting with learnable feature imputation DAGM GCPR 2020
Håkon Hukkelås, Rudolf Mester, Frank Lindseth [PDF] [Github]

DeepPrivacy: A Generative Adversarial Network for Face Anonymization ISVC 2019
(Best paper)
Håkon Hukkelås, Rudolf Mester, Frank Lindseth [PDF] [Github]

Deep Active Learning for Autonomous Perception NIKT 2020
Navjot Singh, Håkon Hukkelås, Frank Lindseth [PDF] [Github]

Autonomous Vehicle Control: End-to-end Learning in Simulated Environments NIKT 2019
Hege Haavaldsen, Max Aasbø, Frank Lindseth, Håkon Hukkelås [PDF]

Talks

DeepPrivacy: Realistic Face Anonymization. August 2021
Given to the UK-based AI interest group brum.ai. [Video]
Introduction to GANs. January 2021
Given at the 2021 Northern Light Deep Learning Workshop. [Video]
DeepPrivacy: Realistic Face Anonymization. January 2021
Given to BRAIN NTNU (student AI organization). [Video]
Improving upon DeepPrivacy for Realistic Image Anonymization. January 2021
Given to the AI Master Thesis Award, NTNU [Video]
DeepPrivacy: Realistic Face Anonymization. October 2020
Given to the personvernkommisjonen appointed by the Norwegian government [Slides]
Image Inpainting With Learnable Feature Imputation. September 2020
Talk at GCPR 2020 [Video]

Open-Source Projects

You can find all my projects on my GitHub page.

Synthesizing Anyone, Anywhere, in Any Pose

Description: A Generative Adversarial Network for synthesizing human figures given a missing region and 17 keypoints indicating the joints of the human body.
Source code: http://github.com/hukkelas/deep_privacy2

DeepPrivacy2: Towards Realistic Full-body Anonymization

Description: DeepPrivacy2 is a toolbox for realistic anonymization of humans, including a face and a full-body anonymizer.
Source code: http://github.com/hukkelas/deep_privacy2

State-of-the-Art Face Detection in Pytorch

Description: A library containing efficient, lightweight, and state-of-the-art face detection models in Pytorch and experimental ports to TensorRT.
Source code: http://github.com/hukkelas/DSFD-Pytorch-Inference

Keypoint Mask R-CNN

Description: A library containing pre-trained, efficient, and high-performing Mask R-CNN models for keypoint and instance segmentation of human figures built on top of Detectron2.
Source code: http://github.com/hukkelas/keypoint_mask_rcnn

DeepPrivacy: A Framework for Realistic Image Anonymization

Description: DeepPrivacy is a fully automatic realistic anonymization framework for faces in images.
Source code: http://github.com/hukkelas/DeepPrivacy