Edge AI Performance on Zephyr Pico vs. Pico 2 misoji　2025/12/8 Zephyr Project Meetup: Toyosu, Tokyo, Japan #ZephyrRTOS 

About Me I am a Hardware and Hobbyist Engineer Handle name：　misoji　@misoji_engineer Blog: The Hardware Guy　(https://misoji-engineer.com/) 

2025/12/8(Mon)　My Seesion ・Debugging Edge AI on Zephyr and Lessons Learned https://ossjapan2025.sched.com/event/29Fm6/ ・Challenging Hardware Contests with Zephyr and Lessons Learned https://ossjapan2025.sched.com/event/29Fmj Agenda An Beginner Introduction to Edge AI on Zephyr https://events.linuxfoundation.org/o pen-source-summit-japan/ Session Continued 

A Beginner-Friendly Approach Edge AI on Zephyr 

Zephyr already provides support ■It’s easy to deploy and run Edge AI models directly on Zephyr. Major Edge AI frameworks already support for Zephyr. https://edgeimpulse.com/ https://www.zephyrproject.org/ Just Build it all together. https://github.com/tensorflow/tflite-micro Zephyr Project We can build together. Edge AI Models 

Edge Impulse ■Overview: ・Easily create AI models (machine learning models) from sensor data. ・Create lightweight models → Integrate them into Zephyr. A development platform for lightweight AI models. https://www.zephyrproject.org/ https://edgeimpulse.com/ 

・Build result example (Motion Recognition) + Good Thing for Zephyr Zephyr & Edge AI fit in small RAM and ROM. ■Including the AI model, fits into kBytes of RAM and ROM.　 Match Low-end SoCs/CPUs ROM:177kB, RAM:27kB 

These were made for hobby, so please use just as a reference. ・pico2-ei-zephyr-demo Raspberry Pi Pico 2(W) & Sensor Board https://github.com/iotengineer22/pico2-ei-zephyr-demo My Test Examples ■GitHub(My Test Example) ・zephyr-ei-xiao-nrf-demo XIAO nRF54L15 Sense https://github.com/iotengineer22/zep hyr-ei-xiao-nrf-demo XIAO nRF54L15 Sense　≒$15 Integrated with ・Accel sensor ・Microphone We can use also Pico/Pico2 + Sensor≒$3 Pico2(W)≒$10 

Raspberry Pi Pico vs. Pico 2 

The Pico2 features upgraded specs. Pico vs. Pico 2 Pico Pico2(W) Feature Raspberry Pi Pico (RP2040) Raspberry Pi Pico 2 (RP2350) Key Differences & Benefits ARM Core Dual Cortex-M0+ Dual Cortex-M33 Significant improvement in processing efficiency (IPC). Clock Speed 133 MHz 150 MHz Higher base clock DSP Extensions None Yes Enables hardware execution of dedicated instructions for signal processing (Filters, FFT, etc.). Upgrade Specs! 

Edge Impulse uses DSP extensions to optimize AI performance. DSP Pico2(W) Pico2(M33) DSP extensions Edge Impulse SDK→CMSIS→DSP https://edgeimpulse.com/ CMSIS・・・Cortex Microcontroller Software Interface Standard DSP(Digital Signal Processing) 　>>Feature Extraction (e.g., Spectrogram, FFT)　 

Benchmark https://youtu.be/Pq3httfiAEs 

Predictions: ・DSP: 39 ms ・Classification: 0 ms ・Anomaly: 0 ms The Pico2 features upgraded specs for Edge AI on Zephyr. Result Pico(M0+) Pico2(M33) About 4 to 5 times faster https://www.zephyrproject.org/ + https://edgeimpulse.com/ *Motion Recoginition Predictions: ・DSP: 190 ms ・Classification: 2 ms ・Anomaly: 1 ms *0ms…Sub-millisecond (can not display) 

Summary 

・Raspberry Pi Pico 2(W) match for debugging Edge AI on Zephyr. ・Lightweight Edge AI model fit Zephyr. (Including the AI model, fits into kBytes of RAM/ROM.) ・If you're interested, please debug it. (Zephyr already provides many support for Edge AI.) Summary I was able to debug Edge AI on Zephyr with Pico/Pico2！ 

In Closing Thank you for your attention.