Cardstat.AI — AI-powered defect detection for trading cards.
Phone photo in, defect map out, in seconds.
Built as our senior capstone at Drexel's College of Computing &
Informatics, Cardstat.AI brings AI to a trading-card grading
process that today is slow, expensive, and inconsistent.
Collectors pay $17 to $1,000 per card and wait weeks or months
for a grade they often can't reproduce. We rebuilt that process
around computer vision.
Under the hood: a custom photometric lighting fixture, normal +
curvature surface maps, a custom pixel-level annotation tool,
and a U-Net + ResNet-18 deep-learning segmentation model trained
on ~800 hand-labeled cards. The result: a working web application
that takes a phone photo and returns a per-pixel defect
breakdown.
Team: Kevin Mastascusa, Chris Jarocha, Asef Ajmain, Dean Huneke,
Hung Phang, Khoi Ma, Jiky Dong.
Drexel CCI Senior Capstone · 2026
