https://yanghu.github.io/tags/yolo/Recent content in Yolo on Yang's NotesHugo -- gohugo.ioen-usyang@yhu.me (Yang Hu)yang@yhu.me (Yang Hu)© 2026 Yang HuThu, 19 Mar 2026 00:00:00 -0800https://yanghu.github.io/posts/frigate-yolov9t-openvino/Thu, 19 Mar 2026 00:00:00 -0800yang@yhu.me (Yang Hu)https://yanghu.github.io/posts/frigate-yolov9t-openvino/<p>Replacing Frigate&rsquo;s default SSD MobileNet detector with YOLOv9t (tiny) running on the Intel N97&rsquo;s integrated GPU via OpenVINO. Covers model export, correct Frigate config, and a critical gotcha that causes 100% false positives if you get it wrong.</p> <h2 class="relative group">Setup <div id="setup" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#setup" aria-label="Anchor">#</a> </span> </h2> <ul> <li><strong>Server:</strong> Intel N97 (Debian 13), 8 camera streams</li> <li><strong>Frigate:</strong> 0.17, running in Docker</li> <li><strong>Detector:</strong> OpenVINO GPU (<code>/dev/dri/renderD128</code>)</li> <li><strong>Previous model:</strong> SSD MobileNet v2 (built-in, 300×300)</li> <li><strong>New model:</strong> YOLOv9t ONNX (320×320, 8.3 MB)</li> </ul> <h2 class="relative group">Why YOLOv9t? <div id="why-yolov9t" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#why-yolov9t" aria-label="Anchor">#</a> </span> </h2> <p>The default SSD MobileNet v2 bundled with Frigate&rsquo;s OpenVINO image is fast and lightweight, but accuracy suffers on partially occluded objects and objects at the edges of the frame. YOLOv9t (tiny) offers meaningfully better detection quality with a similar computational footprint — at 320×320 input and ~18ms inference on the N97 iGPU, it handles 8 concurrent camera streams comfortably.</p>