Kemal Öksüz bio photo

Kemal Öksüz

I'm from a seaside town near Balıkesir, Turkey. Currently in Oxford, UK. Passionate about computer vision.

Twitter   G. Scholar LinkedIn Github e-Mail

Me in a Nutshell

I am a Senior Research Scientist in the Oxford Applied Research Team of Five AI (which has been acquired by Bosch). Our main mission is to help building the technology to enable driverless cars. Towards this aim, I have been closely working with Puneet Dokania to develop accurate and robust computer vision algorithms.

Previously, I received my M.Sc. in 2016 from Dept. of Computer Engineering, Bogazici University under the supervision of Assoc. Prof. Ali Taylan Cemgil and my Ph.D. degree in 2021 from Dept. of Computer Engineering, Middle East Technical University (METU) under the supervision of Assoc. Prof. Sinan Kalkan and Asst. Prof. Emre Akbaş.


Recent News

July 2024 - Our new preprint on understanding safety fine-tuning of Large Language Models is online now: What Makes and Breaks Safety Fine-tuning? A Mechanistic Study

July 2024 - 2 papers accepted to ECCV 2024: (1) On Calibration of Object Detectors: Pitfalls, Evaluation and Baselines with the code at project page.
(2) Bucketed Ranking-based Losses for Efficient Training of Object Detectors. The paper and the code will be made public soon.

October 2023 Our new preprint on zero-shot and weakly-supervised referring image segmentation is online: Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation

September 2023 New preprint proposing an approach to combine various object detectors from different families is online: MoCaE: Mixture of Calibrated Experts Significantly Improves Object Detection

July 2023 1 paper accepted to CVPR 2023: Towards Building Self-Aware Object Detectors via Reliable Uncertainty Quantification and Calibration . Code is available at the project page.

July 2022 - Happy to announce that my PhD Thesis won the best thesis award of Middle East Technical University!