Introduction
This benchmark tests various models (Picovoice Rhino, Google Dialogflow, and Amazon Lex) and provides accuracy using the same dataset. For all input audio files, multiple decibels are provided and the accuracy is calculated. Overall, the benchmark shows that Picovoice Rhino outperforms other models by ~25%. Please refer to the following repo for more information on the benchmark details.
Picovoice Rhino:
# {file_type} {decibel} db:
# {num_inputs} {num errors} {accuracy}
cafe 6 dB:
619 50 0.92
cafe 9 dB:
619 27 0.96
cafe 12 dB:
619 14 0.98
cafe 15 dB:
619 10 0.98
cafe 18 dB:
619 7 0.99
cafe 21 dB:
619 8 0.99
cafe 24 dB:
619 5 0.99
