Joanna Kaleta

I am a PhD researcher in computer vision at Warsaw University of Technology and the Sano Centre for Computational Personalized Medicine, supervised by Prof. Tomasz Trzciński and Dr Marek Kowalski. I am currently a visiting PhD student in the GrUVi Lab at Simon Fraser University, hosted by Prof. Andrea Tagliasacchi. Earlier in 2026, I visited the Computer Vision and Geometry group at ETH Zürich, hosted by Prof. Marc Pollefeys.

I work on 3D spatial representations and generative models to reconstruct, understand, and generate spatially faithful 3D scenes. My research spans feed-forward 3D reconstruction, Gaussian Splatting, neural rendering, diffusion models, and video-based scene representations.

Email / CV / Google Scholar / GitHub / LinkedIn

Portrait of Joanna Kaleta

Publications

Selected publications — the full list is on Google Scholar.

AnyStyle teaser: scenes stylized from text prompts and reference images
AnyStyle: Single-Pass Multimodal Stylization for 3D Gaussian Splatting

Joanna Kaleta, Bartosz Świrta, Kacper Kania, Tomasz Trzciński, Przemysław Spurek, Marek Kowalski

ACM Multimedia 2026

Feed-forward, pose-free 3D reconstruction and zero-shot stylization with text or image conditioning.

project page · paper · code

LumiMotion teaser: a dynamic scene relit under several novel lighting conditions
LumiMotion: Improving Gaussian Relighting with Scene Dynamics

Joanna Kaleta*, Piotr Wójcik*, Kacper Marzol, Tomasz Trzciński, Kacper Kania, Marek Kowalski

CVPR 2026 Highlight

Scene motion as an additional supervisory signal for disentangling material properties from global illumination in Gaussian Splatting reconstructions.

project page · paper · code

LumiGauss teaser: reconstruction, albedo, normals, and novel lighting of an outdoor scene
LumiGauss: Relightable Gaussian Splatting in the Wild

Joanna Kaleta, Kacper Kania, Tomasz Trzciński, Marek Kowalski

WACV 2025 Oral

Relightable Gaussian Splatting from unconstrained photo collections, using precomputed radiance transfer.

paper · code

PR-ENDO teaser: relightable 3D reconstruction of a colon from colonoscopy video
PR-ENDO: Physically Based Relightable Gaussian Splatting for Endoscopy

Joanna Kaleta*, Weronika Smolak-Dyżewska*, Dawid Malarz, Diego Dall'Alba, Przemysław Korzeniowski, Przemysław Spurek

MICCAI 2025

Relightable Gaussian Splatting for colonoscopy, exploiting the co-location of camera and light source.

project page · paper · code

D-MiSo teaser: editable dynamic 3D object represented with multi-Gaussian soup
D-MiSo: Editing Dynamic 3D Scenes Using Multi-Gaussians Soup

Joanna Waczyńska, Piotr Borycki, Joanna Kaleta, Sławomir Tadeja, Przemysław Spurek

NeurIPS 2024

A hierarchical, editable Gaussian Splatting representation for dynamic 3D scenes.

project page · paper · code

Simulator renders translated into realistic endoscopic images
Minimal Data Requirement for Realistic Endoscopic Image Generation with Stable Diffusion

Joanna Kaleta, Diego Dall'Alba, Szymon Płotka, Przemysław Korzeniowski

NeurIPS 2023 Workshop on Diffusion Models NVIDIA Best Student Paper Award

Medical Sim2Real: generating fully annotated, realistic endoscopic data from medical-simulator renders. Extended version published in IJCARS.

paper · code

Research experience

08/2026 – present
Simon Fraser University, GrUVi Lab — Visiting PhD Student, hosted by Prof. Andrea Tagliasacchi. Investigation of spatial understanding in video diffusion models: can a video model re-generate a dynamic video, conditioned on photos of the same scene after spatial changes?
02/2026 – 05/2026
ETH Zürich, Computer Vision and Geometry group — Visiting PhD Student, hosted by Prof. Marc Pollefeys. Test-time refinement of feed-forward 3D reconstruction: recurrent refinement of camera poses and Gaussians. We show that pose refinement is an important factor in boosting reconstruction quality.
10/2023 – present
Warsaw University of Technology — PhD Student, Doctoral School, supervised by Prof. Tomasz Trzciński and Dr Marek Kowalski. 3D reconstruction and neural rendering (Gaussian Splatting, inverse rendering), generative vision models.
10/2022 – present
Sano Centre for Computational Personalized Medicine — PhD Student Researcher. Computer vision and generative modeling for medical imaging, endoscopy, and robotics-oriented applications.

Awards & grants

2026
NAWA scholarship funding the research visit at Simon Fraser University.
2023 – 2027
Researcher in the NCN PRELUDIUM BIS grant “Towards Real-Time Volumetric 3D Reconstruction for Videos”.
2023
NVIDIA Best Student Paper Award — NeurIPS Workshop on Diffusion Models.
2023
1st place, MICCAI PitVis Challenge — surgical instrument recognition in endoscopic pituitary surgery (challenge report).

Teaching & service

Teaching
Rendering and Animation Techniques — lab classes covering Blender fundamentals, 3D modeling, rendering, and animation, with practical exercises on building simple scenes.
Teaching
Introduction to Artificial Neural Networks — labs with mini-lectures on machine learning foundations: MLPs, backpropagation, recurrent networks, generative models, and LLMs.
Reviewing
Reviewer for NeurIPS, CVPR, ICML (Silver Reviewer 2026), ECCV, MICCAI, and WACV.

Industry experience

05/2022 – 09/2022
Appsilon — Machine Learning Engineer. Image classification and object detection in Data4Good wildlife-protection projects; GAN-based human motion style transfer.
10/2020 – 04/2022
RedCompass Labs — Data Scientist. Anomaly detection, graph analysis and NLP on financial transactions to detect modern slavery patterns; Spark pipelines and synthetic transaction data generation.

Education

03/2018 – 06/2021
MSc in Computer Science, Warsaw University of Technology — graduated with distinction.
Erasmus
Exchange semesters at the Polytechnic University of Catalonia and the University of Vienna.