Based on the ODYSSEY-X86J4125 SBC, the Odyssey Blue tiny PC is now available in a package with the Frigate open source NVR platform, which supports real-time local object recognition, and the Coral USB AI accelerator. The Intel Celeron J4125 quad-core Gemini Lake Refresh processor, 8 GB of RAM, and a 128 GB SSD packed with an unidentified Linux OS (likely Debian 11) and Frigate Docker container are all included in the Odyssey Blue tiny PC. When using a Coral USB accelerator, the solution can perform object detection at 100 or more frames per second.
In this article, we’ll concentrate on Frigate NVR because the hardware isn’t really novel and has already been explored in detail.
OpenCV and Tensorflow are used by Frigate, an open source NVR application created for Home Assistant with AI-powered object detection, to carry out real-time object detection locally for IP cameras. Frigate operates as a Docker container. Among the highlights are:
- Makes heavy use of multiprocessing with a focus on real-time rather than processing every frame
- TensorFlow’s object detection operates in distinct processes for maximum FPS, with very minimal overhead motion detection to choose where to perform it.
- Uses MQTT for communication to make it simple to integrate with other systems
- Records video with options for object-based retention.
- 24-hour recording
- Using RTMP to restream to lessen the number of connections to your camera
Although Frigate may conduct object detection using the processor, utilising the Google Coral accelerator greatly enhances performance and allows for over 100 frames per second detection. The manual covers suggested gear, setup guidelines, configuration, Home Assistant integration, and more. In that order, Dahua/Loryta, Hikvision, and Amcrest IP cameras are advised, but Reolink users frequently report issues, thus PoE cameras should be used over WiFi cameras.
Oddly, the kit comes with a USB accelerator rather than the required Odyssey Blue (J4125), which is the only server on the list to support the M.2 Coral AI accelerator. The Minisforum GK41, Beelink GK55, Atomic Pi, and Raspberry Pi 4 are further suggested/tested systems, albeit the latter two are only appropriate for a limited number of cameras.
For $328.99 including delivery, Seeed Studio offers the Frigate NRV bundle kit with Odyssey Blue and the Coral USB AI accelerator. As was already said, it shouldn’t be too difficult to build your own Frigate NVR using just about any modern small PC and a Google Coral AI accelerator, so it should be most intriguing for those who want one that works right out of the box. The latter, however, might not be as straightforward to find, as Seeed Studio informed CNX Software that “considering the very limited supply of Coral USB Accelerator at this stage, only the Frigate NVR bundle kit is available online now,” and that they only intend to resume selling the Coral USB Accelerator separately when the supply shortage is resolved.