In the latest MLPerf Tiny Submission, Skymizer showcased impressive AI capabilities, especially in Anomaly Detection, using the Asus Tinker V with Renesas RZ/Five 1GHz based on Andes RISC-V IP. With Skymizer’s ONNC compiler and Andes AIRE NN library, we achieved significant performance in Anomaly Detection and Image Classification, maintaining high accuracy.

Notably, counting all submitters from 2021 to 2024, Skymizer’s Anomaly Detection ranks in the top 6 above all 61 results, utilizing only a 1GHz RISC-V CPU, without the need for a higher frequency processor or multiple processors, as seen with the other top performers.

This highlights the effectiveness of RISC-V based MPUs in AI applications, particularly in audio-related tasks. We eagerly anticipate exploring more AI opportunities in both MCUs and MPUs, driving innovation in AI technology.

Information:
–MLPerf Tiny v1.2 https://mlcommons.org/2024/04/mlperf-tiny-v1-2-results/

–Skymizer contact: [email protected]