Kacper Kania

PhD Student

Warsaw University of Technology

About Me

I am a 3nd year PhD Student at Warsaw University of Technology. The PhD programme is financed by Microsoft Research Cambridge.

My research revolves around 3D human animation and reconstruction in partial information and low-shot regime. My interests include computer graphics and application of machine learning methods in common 3D algorithms.

Interests

Computer Graphics
Computer Vision
Machine Learning

Education

PhD (ongoing) "Low Shot Realistic Human Rendering from Partial Information"
Warsaw University of Technology
PhD (moved to Warsaw University of Technology) "3D Shape Reconstruction of Objects in the Scene", 2020
Wrocław University of Science and Technology
MSc in Computer Science with Data Science specialization, 2019
Wrocław University of Science and Technology
BSc in Computer Science, 2018
Wrocław University of Science and Technology

Publications

BlendFields: Few-Shot Example-Driven Facial Modeling

We generate realistic 3D avatars from just as few as five multi-view frames. Compared to the previous approaches, which may use up to 1 million images, our approach is data-efficient and easy to train. We show through the experiments, that BlendFields achieves the best results for different faces.

Kacper Kania, Stephan J. Garbin, Andrea Tagliasacchi, Virginia Estellers, Kwang Moo Yi, Julien Valentin, Tomasz Trzciński, Marek Kowalski

PDF Video Project Page

BlendFields: Few-Shot Example-Driven Facial Modeling

CoNeRF: Controllable Neural Radiance Fields

We extend neural 3D representations to allow for intuitive and interpretable user control beyond novel view rendering (i.e. camera control). We demonstrate, to the best of our knowledge, for the first time novel view and novel attribute re-rendering of scenes from a single video.

Kacper Kania, Kwang Moo Yi, Marek Kowalski, Tomasz Trzciński, Andrea Tagliasacchi

PDF Code Video Project Page

CoNeRF: Controllable Neural Radiance Fields

TrajeVAE: Controllable Human Motion Generation from Trajectories

We present a novel method for controllable human animation generation. It is the first method that enables generating sequences from any number of motion trajectories, allowing a fine-grained control over the animation.

Kacper Kania, Marek Kowalski, Tomasz Trzciński

PDF Project Page

TrajeVAE: Controllable Human Motion Generation from Trajectories

Representing Point Clouds with Generative Conditional Invertible Flow Networks

In this paper, we propose a simple yet effective method to represent point clouds as sets of samples drawn from a cloud-specific probability distribution. Our model offers competitive or superior quantitative results on benchmark datasets, while enabling unprecedented capabilities to perform cloud manipulation tasks, such as point cloud registration and regeneration, by a generative network.

Michał Stypułkowski, Kacper Kania, Maciej Zamorski, Maciej Zięba, Tomasz Trzciński, Jan Chorowski

PDF Code DOI

Representing Point Clouds with Generative Conditional Invertible Flow Networks

Modeling 3D Surface Manifolds with a Locally Conditioned Atlas

We introduce a Locally Conditioned Atlas (LoCondA), a framework for representing a 3D object hierarchically in a generative model. We show that the proposed approach provides structurally coherent reconstructions while producing meshes of quality comparable to the competitors.

Przemysław Spurek, Sebastian Winczowski, Maciej Zięba, Tomasz Trzciński, Kacper Kania

PDF

Modeling 3D Surface Manifolds with a Locally Conditioned Atlas

See all publications

Internships

Microsoft Research Cambridge Jul 2022 – Sep 2022

Working on realistic 3D human avatars. I extended VolTeMorph to incorporate more realistic facial details that from performing different expressions. Our results have been published at CVPR 2023 conference. You can read more about it here.

University of British Columbia Jun 2021 – Dec 2021

I have developed a method that enables controlling neural radiance fields from sparse, manual annotations. Both annotations are coarse segmentation masks of the areas you want to control, and leave the neural network to learn how to propagate that information across 3D. The work was published at CVPR 2022. You can read more about that here.

Honours

Reviewer at ICCV 2023

May 2023

Outstanding Reviewer at CVPR 2023

Feb 2023

See certificate

Teaching

Introduction to Artificial Intelligence [PL, WSI-24L-G103, WSI-24L-G104]