Mehmet Aygün

I am a researcher in machine learning and computer vision, with a focus on generative and self-supervised models. My work aims to reduce supervision requirements and improve robustness in real-world settings. I received my Ph.D. from the University of Edinburgh, where I was advised by Oisin Mac Aodha. Before that, I worked with Laura Leal-Taixéat the Technical University of Munich on problems in scene understanding.

CV  /  Google Scholar  /  Bluesky

headshot
News
  • Feb 2025, DepthCues got accepted to CVPR 2025.
  • Feb 2024, SAOR got accepted to CVPR 2024.
  • Aug 2023, I joined Meta AI as a Research Scientist Intern.
  • Jul 2022, Our paper about unsupervised semantic correspondence is accepted to ECCV, 2022.
  • Feb 2021, Our paper titled 4D Panoptic Lidar Segmentation is accepted to CVPR, 2021.
  • Oct 2020, Our paper about 3D shape correspondence is accepted to 3DV, 2020.
  • Jun 2018, I received DAAD scholarship for studying in Germany.
  • Oct 2017, at ICCV 2017, our paper received Honorable Mention Award from TASK-CV workshop.
Research

I am broadly interested in machine learning and computer vision. My work focuses on developing models that can learn fundamental priors and representations with minimal supervision, particularly through generative and self-supervised approaches. I am also interested in improving robustness and generalization in these systems.

filters
filters
Learning Shape, Structure, and Semantics: Self-Supervised Learning with 3D Priors
Mehmet Aygün
University of Edinburgh, PhD Thesis, 2025

This thesis advances computer vision by improving semantic correspondence, unsupervised 3D shape recovery, and self-supervised learning with 3D priors. Together, these methods enhance model robustness and move machine perception closer to human-level understanding..

filters
filters
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
Duolikun Danier, Mehmet Aygün, Changjian Li, and Hakan Bilen, Oisin Mac Aodha
CVPR, 2025

A new benchmark designed to evaluate depth cue understanding, with findings across 20 diverse and representative pre-trained vision models.

filters
filters
SAOR: Single-View Articulated Object Reconstruction
Mehmet Aygün and Oisin Mac Aodha
CVPR, 2024

A new self-supervised approach for estimating shape of highly articulated objects such as animals from single-view images.

filters
filters
Enhancing 2D Representation Learning with a 3D Prior
Mehmet Aygün, Prithviraj Dhar, Zhicheng Yan, Oisin Mac Aodha, and Rakesh Ranjan
CVPR 2024 Workshop on Representation Learning with Very Limited Images

A new approach using 3D reconstruction to enhance the robustness of self-supervised vision models.

filters
filters
Demystifying Unsupervised Semantic Correspondence Estimation
Mehmet Aygün and Oisin Mac Aodha
ECCV, 2022

A novel unsupervised approach for semantic correspondence problem, an evaluation framework along with a metric, and through evaluation of current methods.

filters
filters
4D Panoptic Lidar Segmentation
Mehmet Aygün, Aljosa Osep, Mark Weber, Maxim Maximov, Cyrill Stachniss, Jens Behley and Laura Leal-Taixe
CVPR, 2021

A new method and evaluation metric for the task of 4D panoptic segmentation.

filters
filters
Unsupervised Dense Shape Correspondence using Heat Kernels
Mehmet Aygün, Zorah Lähner and Daniel Cremers
International Conference on 3D Vision (3DV), 2020

An emprical study showing that Heat Kernels can replace Geodesic Matrices for unsupervised dense shape correnspondence problem