Arda Cinar Demirtas

I'm doing my Master's at ETH Zürich in Information Technology and Electrical Engineering, specializing in computer vision and robotics. I also work as a member of technical staff at microagi in Zürich.

At ETH, I'm doing a semester project at the Computer Vision and Geometry Lab with Sunghwan Hong and Marc Pollefeys, adapting a feed-forward 4D reconstruction model (V-DPM) to event cameras by distilling it from paired RGB and event data.

I did my Bachelor's in Electrical and Electronics Engineering at Bilkent University, where I worked on medical image synthesis with Prof. Tolga Çukur. During an exchange at EPFL, I worked on part-based 3D shape representation with Prof. Pascal Fua.

Recent

• 2026

  I2I-Mamba in early access at IEEE Transactions on Biomedical Engineering.

• Sep 26

  Joined microagi as a member of technical staff.

• 2026

  Our 360° visual-inertial SLAM pipeline ranked among the top submissions at the Hilti × Trimble SLAM Challenge, ICRA 2026.

• 2026

  2nd in our track at the Europe Embodied Hackathon in Munich, with a Franka Panda that packs objects into boxes.

• 2025

  PartSDF published in TMLR.

• 2026

  I2I-Mamba: Multi-modal Medical Image Synthesis via Selective State Space Modeling

  Omer F. Atli, Bilal Kabas, Fuat Arslan, Arda C. Demirtas, Mahmut Yurt, Onat Dalmaz, Tolga Çukur. IEEE Transactions on Biomedical Engineering, early access.

  papercode

• 2025

  PartSDF: Part-Based Implicit Neural Representation for Composite 3D Shape Parametrization and Optimization

  Nicolas Talabot, Olivier Clerc, Arda C. Demirtas, Doruk Oner, Pascal Fua. Transactions on Machine Learning Research (TMLR).

  papercode

• 2025

  Multi-Contrast MR Image Synthesis with Episodic State-Space Modeling

  Omer F. Atli, Bilal Kabas, Fuat Arslan, Arda C. Demirtas, Mahmut Yurt, Onat Dalmaz, Tolga Çukur. ISMRM 2025, peer-reviewed oral presentation.

  abstract

• 2026

  Event-based 4D reconstruction

  Semester project at CVG, ETH Zürich. Distilling V-DPM, a feed-forward 4D reconstructor, to run on event camera voxel grids, without ground truth, using LoRA-adapted tokenizers and attention distillation from RGB.

• 2026

  360° visual-inertial mapping on a construction site

  3D Vision course project for the Hilti × Trimble SLAM Challenge (ICRA 2026). A hybrid OpenVINS and COLMAP pipeline for a dual-fisheye camera rig fused with IMU data.

• 2026

  Keyboard-typing robot arm

  Robot Learning course project. An SO-101 arm that types words on a real keyboard by looking at it, trained by imitation learning on teleoperated demonstrations with LeRobot.

• 2026

  Pack anything with a Franka Panda

  Europe Embodied Hackathon, 2nd in our track. Semantically aware packing of objects into boxes, with SAM and Gemma 3 running locally on Intel Panther Lake and ROS 2 for calibration and geometry.

• 2025

  Monocular visual odometry (almost) from scratch

  Mini project for Vision Algorithms for Mobile Robotics (Prof. Davide Scaramuzza). Grid-based features, optical-flow tracking and adaptive landmark management, tested on KITTI, Malaga and Parking.