<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Frigate on Tech Zen Den</title><link>https://www.liu.center/tags/frigate/</link><description>Recent content in Frigate on Tech Zen Den</description><generator>Hugo -- gohugo.io</generator><language>en</language><lastBuildDate>Tue, 25 Apr 2023 14:07:04 -0700</lastBuildDate><atom:link href="https://www.liu.center/tags/frigate/index.xml" rel="self" type="application/rss+xml"/><item><title>How to Use Google Coral USB Accelerator with Frigate NVR on Synology NAS (Docker)</title><link>https://www.liu.center/posts/use-frigate-and-coral-usb-accelerator-in-synology-nas/</link><pubDate>Tue, 25 Apr 2023 14:07:04 -0700</pubDate><guid>https://www.liu.center/posts/use-frigate-and-coral-usb-accelerator-in-synology-nas/</guid><description>
&lt;p&gt;Running &lt;strong&gt;Frigate NVR&lt;/strong&gt; for local AI object detection across multiple security cameras is one of the most rewarding home automation projects. However, running object detection purely on an Intel CPU quickly leads to bottlenecks—inference speeds hover between &lt;strong&gt;80ms to 140ms per frame&lt;/strong&gt;, and CPU utilization can easily pin at 80%–100%.&lt;/p&gt;
&lt;p&gt;The &lt;strong&gt;Google Coral USB Accelerator&lt;/strong&gt; (featuring the Edge TPU coprocessor) completely transforms performance. It offloads real-time machine learning inference, dropping inference latencies down to &lt;strong&gt;8ms–12ms&lt;/strong&gt; while reducing CPU usage to a fraction of its former load.&lt;/p&gt;</description></item></channel></rss>