zh/guides/raspberry-pi/

[giscus[bot]]

zh/guides/raspberry-pi/

通过我们的综合指南了解如何在 Raspberry Pi 上部署Ultralytics YOLO11 。获取性能基准、设置说明和最佳实践。

https://docs.ultralytics.com/zh/guides/raspberry-pi/

[zyt2011]

That is so amazing,my raspberry pi 5 had a new function,I like the Yolo!

[glenn-jocher]

@zyt2011 thank you for your enthusiasm! We're thrilled to see you exploring YOLO11 on Raspberry Pi 5 🍓 For optimal performance, check our Raspberry Pi deployment guide with NCNN conversion tips and benchmarks. Happy building with the YOLO community!

[UltralyticsAssistant]

👋 Hello, thank you for sharing your interest in deploying Ultralytics YOLO 🚀 on Raspberry Pi! It looks like you're referencing the Raspberry Pi deployment guide. We recommend utilizing this guide for detailed instructions, performance benchmarks, and best practices tailored for Raspberry Pi.

If this is a 🐛 Bug Report, please share a minimum reproducible example to help us investigate and resolve the issue more efficiently.

If it's a 💡 Question about deployment or setup, kindly provide more context, such as the Raspberry Pi model, operating system, Python version, and the steps you've followed. Including any logs or errors will also assist us in understanding the situation better.

💻 If you'd like to experiment in other environments, YOLO is compatible across multiple platforms and hardware setups. Try it out on free and powerful GPUs with options like:

• 
• 
• 

Quick Setup

Ensure you're using the latest version of ultralytics by upgrading via:

pip install -U ultralytics

Verify your setup meets the proper requirements, such as Python>=3.8 and PyTorch>=1.8.

Community & Support

Engage with the Ultralytics community through the channel that suits you best:

• Discord for real-time interactions 🎧
• Discourse for in-depth discussions 📝
• Subreddit to exchange ideas with other enthusiasts 👥

Status

If this badge is green, all Ultralytics CI tests are passing, ensuring YOLO compatibility across macOS, Windows, and Ubuntu.

This is an automated response to help get you started 😊. An Ultralytics engineer will follow up soon to further assist!

[Beautifulboy111]

Why does my Raspberry Pi 5 have a high inference latency of 150ms when using NCNN for video recognition, even after overclocking? How can I reduce the inference time?

[zyt2011]

Installing a single Hailo-5 is sufficient.

…

________________________________ 发件人: Beautifulboy111 ***@***.***> 发送时间: 2025年4月22日 14:24 收件人: ultralytics/ultralytics ***@***.***> 抄送: zyt2011 ***@***.***>; Mention ***@***.***> 主题: Re: [ultralytics/ultralytics] zh/guides/raspberry-pi/ (Discussion #19489) Why does my Raspberry Pi 5 have a high inference latency of 150ms when using NCNN for video recognition, even after overclocking? How can I reduce the inference time? ― Reply to this email directly, view it on GitHub<#19489 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BJ7DFR53TAD7QVJNO3TTXJL22XOAJAVCNFSM6AAAAABYEEJCHCVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEOJQGYZTCMI>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[glenn-jocher]

@Beautifulboy111 150ms inference latency with NCNN on Raspberry Pi 5 is higher than expected. Here are some steps to reduce it:

1. Use a smaller YOLO11 model - our benchmarks show YOLO11n is significantly faster than YOLO11s (88ms vs 195ms on Pi 5)

2. Reduce input resolution - try 320x320 instead of 640x640, which can cut inference time by up to 70%

3. Optimize memory allocation by running without GUI (Raspberry Pi OS Lite)

4. Ensure you're using proper cooling for sustained performance when overclocked

5. Consider processing at lower FPS if your application allows it

Regarding @zyt2011's suggestion, while accelerators like Hailo can boost performance, they're a separate investment. I'd recommend optimizing your current setup first using the techniques in our Raspberry Pi guide.

What's your specific use case? This would help us provide more targeted optimization advice.

[Edward-YS]

Hello，I'm wondering if there is Benchmarking Results of Two Different Models (YOLO8n, YOLO8s) Across Ten Formats: PyTorch, TF, ONNX, OpenVINO, TF SavedModel, TF GraphDef, TF Lite, PaddlePaddle, MNN, NCNN.