Michal Neoral

I am a computer vision researcher at the Center for Machine Perception (CMP), Czech Technical University in Prague, in the Visual Recognition Group (VRG) at the Faculty of Electrical Engineering. I finished my Ph.D. there in 2025, after a bachelor’s degree in robotics and a master’s in computer vision at the same faculty.

Research interests

Most of my work so far has been about motion: what moves in a video, and where every pixel goes. That covers optical flow with explicit occlusion reasoning, discovery and segmentation of independently moving objects seen from a moving camera, and long-term dense tracking, where correspondences have to survive hundreds of frames rather than two. My dissertation, Dense Motion Estimation in a Monocular Video, collects that line of work.

More recently I have been working on detection, classification and fine-grained recognition: object detection and classification in large video archives with fast incremental learning of new classes, open-set recognition where unknown classes appear at inference time, and compact image descriptors distilled from frozen vision-language models.

Much of this work was done in a long collaboration between CTU and Toyota Motor Europe, which ran from 2015 to 2025 and produced most of my papers and four patent filings.

Education

All three degrees are from the Faculty of Electrical Engineering, Czech Technical University in Prague.

  • Doctoral degree, Ph.D. (Philosophiæ Doctor), 2018 – 2025
    field of study: Informatics / Artificial Intelligence and Biocybernetics
    dissertation: Dense Motion Estimation in a Monocular Video
    supervisor: prof. Ing. Jiří Matas, Ph.D., co-supervisor: Mgr. Jan Šochman, Ph.D.

  • Master's degree, Ing. (MSc equivalent), 2014 – 2017
    programme: Open Informatics, specialization: Computer Vision and Image Processing
    thesis: Object Scene Flow in Video Sequences
    supervisor: Mgr. Jan Šochman, Ph.D.
    awarded the Dean's Prize for an outstanding master thesis

  • Bachelor's degree, Bc. (BSc equivalent), 2011 – 2014
    programme: Cybernetics and Robotics, specialization: Robotics
    thesis: Extraction of Features from Moving Garment
    supervisor: Ing. Pavel Krsek, Ph.D.

Teaching

I have been teaching labs at the faculty since 2018, in Pattern Recognition and Machine Learning, Computer Vision Methods, and Digital Photography Processing. Details are under Teaching, and the rest of the record is in my CV.

Contact