Luka Čehovin Zajc, PhD

Associate Professor
  lukacu
  +386 1 479 8252

About me

I am an associate professor at the Visual Cognitive Systems Laboratory at the Faculty of Computer and Information Science, University of Ljubljana. I teach courses in multimedia systems, robotics, and human-computer interaction.

My current research focuses on robust and self-supervised learning for Earth observation, particularly learning from limited or noisy annotations, multispectral data, satellite image time series, and change detection. I am interested in methods that translate into practical tools for environmental monitoring and resource management.

I am also a founding member and co-organizer of the Visual Object Tracking (VOT) Challenge, where I work on reproducible benchmarking and systematic evaluation of visual trackers. I develop and maintain the open-source VOT evaluation toolkit. In 2026, the VOT Challenge Series received the PAMI Mark Everingham Prize for its long-standing contribution to the computer vision community.

I currently lead the ARIS-funded RoDEO project, which studies robust deep learning for Earth observation. I also contribute to research on complex object motion and cultivated-terrace mapping. Across these areas, I value open-source software, reproducible research, and collaboration across disciplines.

You can find my research record on Google Scholar and ORCID, and my non-scientific work on my personal portfolio.

Active research

Remote sensing

Contains 4 subtopics
We use modern computer vision and machine learning methods to analyze the growing volume of satellite and aerial imagery and address a range of remote-sensing problems.

Remote sensing change detection

We develop supervised and unsupervised methods for detecting semantic changes in multi-temporal Earth-observation imagery, with an emphasis on robust design, generalization, and limited annotations.

Foundation models for Earth observation

We study foundation models for Earth observation, focusing on transferable representations, multispectral imagery, cross-modal knowledge transfer, and efficient adaptation across sensors and tasks.

Active projects

RoDEORobust Deep Learning for Earth Observation​

January 2025 - December 2027
This ARIS funded project investigats the relationship between sensor fusion and self-supervised learning for data-driven Earth Observation. We focus on the role of self-supervised deep learning for sensor fusion from the perspective of different sources with different spatial resolutions and spectral coverage. The project is grounded in a real-world application in the field of hydrology, where the goal is to predict the water level in rivers using satellite and drone imagery.

Geospatial Information Technologies for a Resilient and Sustainable Society

July 2025 - June 2028
The GeoAI project develops advanced geospatial modeling and analytics methods to support the sustainable management of the built and natural environment.

COMETAdvanced Computer Vision for Understanding Complex Object Motion in Dynamic Environments

January 2025 - December 2027
This project aims to develop a novel motion understanding paradigm, centered on automatically determining the minimal scene understanding required to track one or multiple objects throughout a video. It tackles three core challenges: developing a few-shot object detector capable of identifying all objects in a category based on limited examples, tracking individual objects amid distractors, and extending this to track transformable objects in complex environments.

The Life and Death of Cultivated Terraces: Computer-Based Recognition and Spatial Analysis of Terraces

January 2025 - December 2027
Terraced landscapes are cultural landscapes of special significance. Cultivated terraces, which are often intended for farming, ensure food and have a priceless ecological, cultural, and historical value. Terraced landscapes are disappearing in places due to overgrowth or improper maintenance. In order to raise awareness of their significance, the Honghe Declaration was passed at the global level in 2010. Slovenia still has not managed to implement suitable criteria for terrace identification or introduce a comprehensive management system despite having officially recognized terraces as a valuable element for landscape preservation (see the Decree on Cross-Compliance, Official Gazette of the Republic of Slovenia, no. 97/2015). A precise overview (register) of the locations of cultivated terraces and spatial analyses are required as the foundation for any further measures.

Recent publications