Serdar Anıl
Demirbaş
I build computer vision that has to run on a device, not in a notebook.
Since 2021 I have worked on real-time image processing in the defence industry: target detection in C++, tracking in Python, operator interfaces in Qt, all of it integrated on NVIDIA Jetson. Alongside that I lead a five-engineer software team at Atılım UAV and build ball and player tracking for a sports technology client. The thread through all of it is the same: turning an image into an action, fast enough and small enough to fit on real hardware.
Work
Autonomous UAS Image Processing and Mission Software
Real-Time Target Detection and Tracking
Sports Field Video Analytics
Writing
Thermal imaging for detection work: what changes when light becomes heat
A thermal camera does not see light, it sees temperature. That breaks half of your RGB intuition. What I had to relearn to do detection on the thermal channel.
The gap between the notebook and the board
A detector that scores well on a workstation and fails on an embedded target is usually not a model problem. Four places the gap actually opens up.
Why I labelled it a shot candidate and not a goal
Building event detection from ball trajectory on uncalibrated footage, and why naming the limitation in the output was cheaper than pretending it was not there.
Get in touch
I am based in Ankara and open to conversations about real-time computer vision and embedded imaging work.