Hardware recommendation for my setup

[c0dingDom]

Hey guys,

I've got 6 cameras around my house:

3x Reolink RLC-810A
2x Reolink RLC-811A
1x Reolink E1 Zoom(4K) - as Pet Cam (we can ignore it, I guess)

All cameras, but the pet cam, are wired with cat in a gigabit network. I am pretty interested in frigate and want to give it a shot, but I am not sure, which hardware I shall use to run it smoothly with 5 cameras on max resolution.

I recently got a ZimaBoard 2 1664, but it's not installed yet. Initial plan was to use it as replacement for my old home server with debian, but I guess I could use it for frigate.

Is it enough? Would a Raspberry Pi 5 with AI detector hat maybe a better product? This way I could use a ZimaBoard as homeserver as intended.

Would be thankful for recommendations.

Thanks in advance.

[inkeep-ai-assistant[bot]]

Based on your camera setup, here's what you need to know:

Your Cameras

Your Reolink RLC-810A and RLC-811A cameras are 8MP (4K) cameras, which fall into the "6MP or higher, Older" category(1). For these cameras, RTSP is the recommended stream type(1).

Hardware Requirements

ZimaBoard 2 1664:
The ZimaBoard 2 1664 has an Intel N100 CPU with integrated graphics. According to the documentation, the Intel N100 can handle object detection with approximately 15ms inference time for MobileNetV2 and around 25-30ms for YOLO-NAS at 320 resolution(2). However, it can only run one detector instance(2).

With 5 high-resolution cameras (8MP), the N100's single detector instance may struggle to keep up, especially during high activity periods. The CPU will also need to handle video decoding for all streams.

Raspberry Pi 5 with Hailo AI Hat:
This would be a stronger option. The Hailo-8L (available as a HAT for Raspberry Pi) offers approximately 10-11ms inference time for YOLOv6n(3). The Hailo detector supports multiple model architectures and runs efficiently(2). The Pi 5 setup would handle your camera load better than the N100.

Recommendation

For your 5 cameras at maximum resolution:

1. Use the Raspberry Pi 5 with Hailo-8L HAT for Frigate - this provides dedicated AI acceleration and can handle your camera count more effectively(3)(2)
2. Keep the ZimaBoard as your home server as originally intended
3. Configure your Reolink cameras to use RTSP streams(1) with the http-flv protocol for better stability where possible

The N100 could work, but you'd likely need to reduce detection resolution or frame rates to maintain performance with 5 high-resolution cameras.

📚 Sources:

1. Camera Specific Configurations
2. Supported Hardware
3. Recommended hardware

---

Have another question? Just tag @inkeep.