{"slug": "frigate-with-hailo-for-object-detection-on-a-raspberry-pi", "title": "Frigate with Hailo for object detection on a Raspberry Pi", "summary": "This article provides a guide for configuring the Frigate security camera software to use a Hailo AI coprocessor (such as the Hailo-8 or Hailo-8L on a Raspberry Pi 5) for object detection. The author details the installation process and a specific fix for a \"force_desc_page_size\" error, which involves creating a kernel module configuration file. The guide notes that the cheaper Hailo-8L offers a cost-effective alternative to the discontinued Google Coral TPU, achieving low-latency inference (around 12ms) with minimal CPU usage.", "body_md": "Frigate with Hailo for object detection on a Raspberry Pi\nI run Frigate to record security cameras and detect people, cars, and animals when in view. My current Frigate server runs on a Raspberry Pi CM4 and a Coral TPU plugged in via USB.\nRaspberry Pi offers multiple AI HAT+'s for the Raspberry Pi 5 with built-in Hailo-8 or Hailo-8L AI coprocessors, and they're useful for low-power inference (like for image object detection) on the Pi. Hailo coprocessors can be used with other SBCs and computers too, if you buy an M.2 version.\nFrigate offers Hailo support, but getting it working on my new build (pictured above—full writeup coming soon) threw me for a loop, so I'm documenting the process for my build here, for my own future reference.\nAssuming you have a fresh Pi OS install on a Pi 5 or CM5, and you have the Hailo module connected via PCIe (either a HAT+ module, or via an M.2 slot), do the following:\n- Follow Frigate's guide to install the Hailo-8 driver from source.\n- Follow Frigate's guide for setting up Hailo as an object detector in your Frigate\nconfig.yml\n(note: I use my open source Pi NVR project). - Start Frigate.\nIf you see a message like:\nWARNING : CPU detectors are not recommended and should only be used for testing or for trial purposes.\nThen your detector configuration is likely wrong. Check back through your model\nand detectors\nconfig in frigate.yml\nand try again :)\nPCIe force_desc_page_size\nissue\nOutside of configuration issues that took a few tries to fix, I ran into this error message when running docker logs frigate\n:\n[HailoRT] [error] CHECK failed - max_desc_page_size given 16384 is bigger than hw max desc page size 4096\n...\nRuntimeError: HailoRT inference thread has stopped, restart required.\nThat error led me to this forum thread, which suggested a fix:\nCreate a file /etc/modprobe.d/hailo_pci.conf\n(e.g. with sudo nano\n), with the following contents:\noptions hailo_pci force_desc_page_size=4096\nThen either reboot the Pi, or run the following commands to unload, then reload, the hailo driver:\nsudo modprobe -r hailo_pci\nsudo modprobe hailo_pci\nThe Frigate container logs should report the Hailo detector is now in use:\n[2026-02-18 14:14:35] detector.hailo INFO: Starting detection process: 621\nAnd now, the system dashboard is showing hailo\nas running at around 12ms inference speed, with very low CPU usage overall (this was testing with two cameras):\nThe nice thing is, this works on the cheaper Hailo-8L that you can find in the $70 AI HAT+ (the 'base model'), which is often much cheaper than the now-ancient Google Coral TPU—at least the USB version that's more broadly compatible.\nI'm still at a loss as to why Google abandoned Coral after one generation. It seems like they could've dominated the 'edge TPU' market, but they kinda gave it up after hitting 4 TOPS.", "url": "https://wpnews.pro/news/frigate-with-hailo-for-object-detection-on-a-raspberry-pi", "canonical_source": "https://www.jeffgeerling.com/blog/2026/frigate-with-hailo-for-object-detection-on-a-raspberry-pi/", "published_at": "2026-02-18 20:50:00+00:00", "updated_at": "2026-05-22 21:42:21.713032+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "open-source", "hardware", "developer-tools"], "entities": ["Frigate", "Hailo", "Raspberry Pi", "Coral TPU", "Pi NVR", "Hailo-8", "Hailo-8L", "CM4"], "alternates": {"html": "https://wpnews.pro/news/frigate-with-hailo-for-object-detection-on-a-raspberry-pi", "markdown": "https://wpnews.pro/news/frigate-with-hailo-for-object-detection-on-a-raspberry-pi.md", "text": "https://wpnews.pro/news/frigate-with-hailo-for-object-detection-on-a-raspberry-pi.txt", "jsonld": "https://wpnews.pro/news/frigate-with-hailo-for-object-detection-on-a-raspberry-pi.jsonld"}}