Semih Günel

I am a Machine Learning Engineer at the Meta London office working on vision based Generative AI tools for Monetization. I am interested in computer vision, machine learning and self-supervised learning.

Previously, I graduated with my Ph.D. at EPFL, Lausanne, Switzerland in 2022. I did my Ph.D. in Computer Vision Lab under the supervision of Pascal Fua. My first author work was published in journals such as IJCV and Nature Methods, and workshops in NeurIPS, ICCV, and CVPR.

I obtained my B.Sc. in Computer Science at Bilkent University, Ankara, Turkey.

Email  /  Google Scholar  /  Linkedin  /  Twitter  /  Github

profile photo
Main Publications
Overcoming the Domain Gap in Contrastive Learning of Neural Action Representations
Paper  /  Paper (Workshop)  /  Video  /  Code  /  Dataset  /  Bibtex
Semih Günel, Florian Aymanns, Sina Honari, Pavan Ramdya, Pascal Fua
IJCV, 2022 / NeurIPS Workshop, 2021

Self-supervised action recognition and domain adaptation on human and animal neural recordings using 3D markerless motion capture.

LiftPose3D, a deep learning-based approach for transforming 2D to 3D pose in laboratory animals
Paper  /  Video  /  Code  /  Dataset  /  Bibtex
Adam Gostztolai*, Semih Günel*, Victor Lobato Rios, Marco Pietro Abrate, Daniel Morales, Helge Rhodin, Pascal Fua, Pavan Ramdya
Nature Methods, 2021 / CVPR Workshop, 2021

Monocular 3D pose estimation on laboratory animals.

Deformation-aware Unpaired Image Translation for Pose Estimation on Laboratory Animals
Paper  /  Poster  /  Video  /  Code  /  Dataset  /  Bibtex
Siyuan Li, Semih Günel, Mirela Ostrek, Pavan Ramdya, Pascal Fua, Helge Rhodin
CVPR, 2020

Unsupervised pose-estimation using rough animal models.

DeepFly3D, a deep learning-based approach for 3D limb and appendage tracking in tethered, adult Drosophila
Paper  /  Video  /  Code  /  Dataset  /  Bibtex
Semih Günel, Helge Rhodin, Daniel Morales, João Campagnolo, Pavan Ramdya, Pascal Fua
eLife, 2019

Multi-view motion capture system for the model organism Drosophila Melonagaster.

Includes a dataset with 1M images containing 38 landmark annotations. Mentioned In:
Gravity as a Reference for Estimating a Person's Height from Video
Paper  /  Poster  /  Video  /  Dataset  /  Bibtex
Didier Bieler, Semih Günel, Pascal Fua, Helge Rhodin
ICCV, 2019

Unsupervised monocular scale estimation on videos using object free-fall motion.

What Face and Body Shapes Tell Us About Height
Paper  /  Poster  /  Dataset  /  Bibtex
Semih Günel, Helge Rhodin, Pascal Fua
ICCV Workshop, 2019  

Monocular scale/height estimation using single images. Monocular human pose dataset with 1e6 images and 12,000 unique people with known height.

Vision Consultation
Long-term imaging of the ventral nerve cord
Paper  /  Code  /  Dataset  /  Bibtex
Laura Hermans*, Murat Kaynak*, Jonas Braun, Victor Lobato Ríos, Chin-Lin Chen, Semih Günel, Florian Aymanns, Mahmut Selman Sakar, Pavan Ramdya
Nature Communications, 2022  

Long-term microscopy imaging of animal nervous system.

Ascending neurons convey behavioral state to integrative sensory and action selection centers in the brain
Paper  /  Video  /  Code  /  Dataset  /  Bibtex
Chin-Lin Chen, Florian Aymanns, Ryo Minegishi, Victor D. V. Matsuda, Nicolas Talabot, Semih Günel, Barry J. Dickson, Pavan Ramdya
Nature Neuroscience, 2023  

Understanding the behavioral properties of neurons using two-photon recordings and motion capture.

JB's HTML template.