Can Sarı

I'm a senior Computer Engineering student at Bahçeşehir University in Istanbul, where I do research in 3D computer vision at the BAU Vision Lab with Barış Özcan. I work on point cloud understanding, mainly keypoint and corner detection, and I'm increasingly interested in neural rendering with 3D Gaussian Splatting.

I spent spring 2026 at TU Berlin on an Erasmus exchange, and I graduate in June 2027. I'm open to research engineering roles and internships in the meantime.

Email/ Scholar/ GitHub/ LinkedIn

Research

I'm interested in how networks can reason about 3D structure directly: finding geometrically meaningful points in raw point clouds, and representing scenes in ways that are both accurate and fast to render.

First page of the Learned Suppression paper

Learned Suppression for 3D Keypoint Detection with a Graph-Transformer Backbone

Batuhan Arda Bekar, Can Sarı, Hüseyin Can Gülkan, Barış Özcan

ACCV, 2026

project page paper arXiv supplementary code

A small learned graph network replaces NMS and DBSCAN at the end of 3D keypoint detectors, deciding which candidates to keep and moving them onto the right location. Best Corner F1 on Building3D Entry-Level, and a single model beats per-category KeypointDETR on 12 of 16 KeypointNet categories.

bibtex

@misc{bekar2026learnedsuppression3dkeypoint,
      title={Learned Suppression for 3D Keypoint Detection with a Graph-Transformer Backbone},
      author={Batuhan Arda Bekar and Can Sarı and Hüseyin Can Gülkan and Barış Özcan},
      year={2026},
      eprint={2605.15088},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2605.15088},
}

Current work

Undergraduate researcher, BAU Vision Lab, Istanbul

Jun 2025 – now

3D reconstruction, advised by Asst. Prof. Barış Özcan.

Industry

IT Applications Intern, KPMG Turkey

Mar – Aug 2025

Built an Azure DevOps CI/CD pipeline (YAML, PowerShell) that replaced a fully manual release process for internal apps, and coordinated a rollout across engineering, legal, and audit teams.

Software Development Intern, Sisal Digital Hub Turkey

Jul – Aug 2024

Built and shipped GitPeek, a React Native and Spring Boot mobile app, from scratch in four weeks, owning the API, UI, build pipeline, and tests.

Projects

Federated handwriting recognition, TU Berlin

2026

A CNN trained across four clients on FEMNIST with Flower and FedAvg (86% validation accuracy), tested under non-IID splits and deployed on Google Kubernetes Engine with live dashboards of client contributions.

Hybrid recommender system

2025

Matrix factorization combined with neural collaborative filtering, written in C++. Ranked 1st of 300 students in a university programming competition.

Education

B.Sc. Computer Engineering, Bahçeşehir University

2023 – 2027

Erasmus exchange, Technische Universität Berlin

Mar – Jul 2026

Deep Learning 2, Foundations of Data Science & AI, Federated Machine Learning.