Tuna Gül

AI / Hackathons

Holdpoint In progress

A concrete crack-inspection tool: it finds cracks in a photo, gives their width as a millimetre band using a scale card, and leaves the decision to a human inspector.

Role
Directed AI coding agents
Period
Oct 2026
Team
Solo
Stack
Python · PyTorch · OpenCV · Streamlit
Impact
Built during AMD Developer Hackathon: ACT III (not submitted)

Code ↗

Holdpoint

What I did

During the AMD Developer Hackathon: ACT III I directed AI coding agents building a concrete crack-inspection tool (work in progress). A UNet segments the cracks, a printed ArUco card gives the scale, and the tool reports crack width as a millimetre band rather than one false-precise number (the card step is not yet tested on real phone photos). If there is no card, it says the width cannot be measured. The final decision always stays with a human inspector.

Gallery

The inspection screen.
The inspection screen.
A sample photo with the printed scale card.
A sample photo with the printed scale card.