The benchmark produces a Computer Vision score. Each sub-test also produces a score. Higher scores indicate better performance.

Computer Vision score    =    1,000,000 × geomean(1/R1, 1/R2, 1/R3)

Where:
R1    =    The result of the TensorFlow test
R2    =    The result of the ZXing test
R3    =    The result of the Tesseract test