Profile picture

Jonáš Kulhánek

https://jkulhanek.com
[email protected] | +420 728 042 412
PhD student
I am a PhD student in the Computer Science doctoral study program at the Department of Cybernetics, Faculty of Electrical Engineering (FEE), Czech Technical University (CTU). I am currently researching the topic of implicit 3D scene representation under the supervision of Torsten Sattler, co-supervised by Marc Pollefeys. The PhD is part of the European Laboratory for Learning and Intelligent Systems (ELLIS) program. I received my BSc degree from FEE, CTU, and my MSc degree from the Faculty of Mathematics and Physics, Charles University, both in computer science. Before starting my doctoral studies, I was a junior researcher at the Czech Institute of Informatics, Robotics and Cybernetics (CIIRC), CTU, where I focused on visual navigation using deep reinforcement learning and neural scene representation and rendering.

News

08/2023 Our Tetra-NeRF paper got accepted to ICCV 2023
01/2023 I joined the ELLIS program with topic: Towards a Unified 3D Scene Representation
11/2022 I finished my internship at Meta Reality Labs
07/2022 Our ViewFormer paper got accepted to ECCV 2022
09/2021 Our dialogue system paper got accepted to EMNLP, 3rd NLP4ConvAI
09/2021 I was a member of the IROS 2021 conference organizing comittee
09/2021 Started PhD at CTU under the supervision of Torsten Sattler
02/2021 Our visual navigation paper got accepted to ICRA 2021, RA-L

Publications

${alt}
WildGaussians: 3D Gaussian Splatting in the WildJonas Kulhanek, Songyou Peng, Zuzana Kukelova, Marc Pollefeys, and Torsten SattlerarXiv
${alt}
NerfBaselines: Consistent and Reproducible Evaluation of Novel View Synthesis MethodsJonas Kulhanek and Torsten SattlerarXiv
${alt}
Dynamic 3D Gaussian Fields for Urban AreasTobias Fischer, Jonas Kulhanek, Samuel Rota Bulò, Lorenzo Porzi, Marc Pollefeys, and Peter KontschiederarXiv
${alt}
Tetra-NeRF: Representing Neural Radiance Fields Using TetrahedraJonas Kulhanek and Torsten SattlerICCV 2023
${alt}
ViewFormer: NeRF-free Neural Rendering from Few Images Using TransformersJonáš Kulhánek and Erik Derner and Torsten Sattler and Robert BabuškaECCV 2022
${alt}
SymFormer: End-to-end symbolic regression using transformer-based architectureMartin Vastl and Jonáš Kulhánek and Jiří Kubalík and Erik Derner and Robert BabuškaIEEE Access 2024
${alt}
AuGPT: Auxiliary Tasks and Data Augmentation for End-To-End Dialogue with Pre-Trained Language ModelsJonáš Kulhánek and Vojtěch Hudeček and Tomáš Nekvinda and Ondřej DušekNLP4ConvAI, EMNLP 2021
${alt}
Visual Navigation in Real-World Indoor Environments Using End-to-End Deep Reinforcement LearningJonáš Kulhánek and Erik Derner and Robert BabuškaRA-L 2021
${alt}
Vision-based Navigation Using Deep Reinforcement LearningJonáš Kulhánek and Erik Derner and Tim de Bruin and Robert BabuškaECCV 2019

Teaching

Reading Group on Implicit Scene RepresentationSpring 2023
CTU in Prague
NeRFs, surface reconstruction, volumetric rendering, etc.
Introduction to Artificial IntelligenceSpring 2020
Teaching assistant at FEE, CTU in Prague
Searching state space, two player zero-sum games, constraint satisfaction problems, planning, knowledge representation in FOL, decision making with imperfect information, MDP, evolutionary techniques

Education

Charles University(Jun 2021)
Master in Artificial Intelligence
Faculty of Mathematics and Physics, graduated summa cum laude
Czech Technical University in Prague(Jun 2019)
Bachelor in Computer Science
Faculty of Electrical Engineering, graduated summa cum laude

Work experienceSelected experience

Czech Institute of Informatics, Robotics and Cybernetics, CTUSep 2021 — Present
PhD candidate
  • Supervisor Torsten Sattler
  • Using priors in neural rendering and neural scene representation
Supervisors Torsten Sattler and Marc Pollefeys, 3D scene reconstruction
Czech Institute of Informatics, Robotics and Cybernetics, CTUAug 2020 — Aug 2021
Junior researcher
  • Visual navigation using deep reinforcement learning (DRL)
  • Implementing SoTA DRL algorithms including Rainbow, PAAC, AlphaZero
  • Publishing an open-source 3D environment simulator to train DRL agents
  • Working with Tomas Mikolov on cellular automata applied to NLP
  • Researching dialogue systems with pre-trained LMs (GPT2, BERT)
  • Running large distributed GPU training
  • Achieving 3rd place in the DSTC9 end-to-end multi-domain dialogue task
  • PyTorch, Slurm, Numpy, ROS, ...
Visual navigation using deep reinforcement learning, end-to-end dialogue systems using transformer models
Delft University of TechnologyFeb 2019 — Apr 2019
Research intern
  • Visual navigation using deep reinforcement learning (DRL)
  • Working under Prof. Dr. Robert Babuska at 3me
  • Publishing open-source DRL PyTorch library
  • Evaluating navigation algorithms on real mobile robots
  • PyTorch, Slurm, Numpy, ROS, ...
Working with Robert Babuska on visual navigation using deep reinforcement learning
ValatronJun 2018 — Jan 2019
Backend developer
  • Designing Node.js API backend and .NET Core backend
  • Leading a team of developers
  • Working on SPA frontend (React.js)
  • Deploying applications built using Azure Functions
  • Node.js, .NET Core, Azure Functions, React.js, ...
Center for Machine PerceptionFeb 2018 — Jun 2018
Research intern
  • Building object detectors using fully convolutional neural networks
  • Implementing RetinaNet, FasterRCNN, MaskNet (SoTA at that time)
  • Preparing object detection datasets and data pipelines
  • TensorFlow, Numpy, OpenCV, ...
Object detection for automotive industry
Alza.czJun 2016 — Feb 2018
Full-stack developer
  • Architecting backend (.NET core, WCF, SOA, SQL)
  • Designing continuous integration in TFS
  • Front-end website development (ASP.NET Web Forms)
CZECH NEWS CENTERMar 2015 — May 2016
Mobile application developer
  • Tydenik Reflex app for Windows Phone — MVVM Light
  • Reflex app for Windows 8 — MVVM Light
  • Blesk app for Windows 8 and Windows Phone 8 — build with Prism
  • .NET, MVVM Light, Microsoft Prism, sqlite3, XAML