This repository contains the school project where we used GPGPU to detect barcode.
End to end application real-time barcode detector.
The inefficient but nice to parallelise algorithm is separated in multiple steps:
- Compute local binary pattern of input image
- Apply a KNN classifier to extract a class represenigs barcode patterns (ie. a database)
- Parallelize using CUDA the local binary pattern and the neareset neighbour search (ie. finding the appropriate class)
- Post-process using mathematical morphology
The python folder contains the ipynb notebook that allows the KNN database generation from images The c++ folder contains a base reference using CPU only The cuda folder contains the parallelized algorithm and benchmarks