May 12, 2021 17:14:00

'Geekbench ML' to measure whether the device is suitable for machine learning apps

'Geekbench ML ', which measures how well a terminal can perform a machine learning application, has appeared in

'Geekbench', one of the standard benchmark software used in PCs and smartphones.

Geekbench ML --Cross-Platform Benchmark
https://www.geekbench.com/ml/

Geekbench ML is

available for Android and iOS . This time I will try the Android version on Pixel 5.

First, install the app on Google Play.

A dialog like this will be displayed at the first startup, so tap 'ACCEPT'.

The benchmark framework looks like a pull-down menu in 'TensorFlow Lite', but you can't choose anything else. You can choose from three backends: 'CPU', 'GPU', and '

NNAPI'.

When I actually benchmarked it, the CPU score was '250'.

GPU is '653'

NNAPI is '414'

You can check how much the score is compared to other terminals with Geekbench Browser.

Geekbench ML Results --Geekbench Browser
https://browser.geekbench.com/ml/v0/inference/