Hello mentors,
I’m interested in the QR/Barcode/ArUco deep detector project for GSoC 2026.
I tested the existing ONNX-based deep detector using OpenCV DNN and verified
the input preprocessing (1×3×640×640) and multi-scale outputs.
When testing on clean QR images, I observed low confidence compared to
OpenCV’s classical QRCodeDetector, which suggests room for improving
generalization.
I’m planning to prototype a lightweight YOLO-style detector, export to ONNX,
and evaluate it using OpenCV DNN before extending to Barcode and ArUco.
I’d appreciate your guidance on whether this approach aligns with the
project expectations.
Thank you.