CoBRA: Controlled Rendering and Annotation of Barcode Images Applied to the Cylindrical Surfaces

QR Codes seem to be the most popular form of machine readable data representation. However, their presence on cylindrical surfaces, like bottles and cans, often complicates scanning. This paper introduces the technology for 3D rendering cylindrical surfaces with barcodes and generating their precise geometry annotations. This technology was implemented in the CoBRA system. It also presents the CoBRA-CYL-QR dataset created using this system. We compare five open-source barcode scanners ( ZBar , ZXing , Quirc , OpenCV , WeChat ) on this dataset, estimating the maximum QR Code curvature allowing for successful decoding. It is shown that barcode scanners fail to decode QR Codes with even slight curvature. The proposed CoBRA dataset synthesis system may be adapted to work with various barcode symbologies or other surface types.