Computer vision researchers build AI app for Android phones

Avatar By Le Williams | 2 years ago

ETH Zurich computer vision researchers have enhanced AI development on smartphones by creating a benchmark system for assessing the performance of several major neural network architectures used for common AI tasks.

The researchers have initiatives to support the progress of chipmakers, in addition to other AI scientists, through the advancement of procuring competitive insights.

With the enhanced AI development, Android developers will have the ability to analyze how fast their AI models will run on different devices. Smartphone techs will be supported by showing whether or not a particular device contains the necessary drivers for AI accelerators.

The app, called AI Benchmark, is available for download on Google Play and can run on any device with Android 4.1 or higher, generating a score the researchers describe as a “final verdict” of the device’s AI performance, as first noted by Tech Crunch.

AI tasks being assessed by their benchmark system include image classification, face recognition, image deblurring, image super-resolution, photo enhancement or segmentation.

Numerous algorithms are in the testing process, which are used in autonomous driving systems.

For Samsung processors, the researchers flag up that all the company’s devices are still running Android 8.0 but AI acceleration is only available starting from Android 8.1 and above. Natch.

They also found CPU performance could “vary quite significantly” — up to 50% on the same Samsung device — because of throttling and power optimization logic. Which would then have a knock on impact on AI performance.

“The researchers hope their score will become a universally accepted metric — similar to DxOMark that is used for evaluating camera performance — and all algorithms included in the benchmark are open source. The current ranking of different smartphones and mobile processors is available on the project’s webpage,” Tech Crunch states.

The benchmark system and app was around three months in development, says AI researcher and developer Andrey Ignatov.