[Support]: Improve Object Detection - Frigate+?

[broyuken]

Describe the problem you are having

I have a couple of floodlight cameras that are mounted fairly high up, and as such the angle that it sees things at may not be typical. I believe because of this, my dogs keep getting detected as people.

Is there something I can change to help frigate detect them as dogs better? I don't want to change the minimum score, because that seems to be fairly accurate for actual people however dogs seems to be less reliable. Is there a config that says that this is a higher up view, or would frigate+ help out with this at all? the AI summary seems to know what's going on even if Frigate doesn't.

Version

0.16.0-c2f8de9

What browser(s) are you using?

Version 141.0.7390.55 (Official Build) (64-bit)

Frigate config file

N/A

Relevant Frigate log output

N/A

Relevant go2rtc log output

N/A

FFprobe output from your camera

N/A

Frigate stats

N/A

Install method

Docker Compose

docker-compose file or Docker CLI command

N/A

Object Detector

Coral

Network connection

Wired

Camera make and model

Reolink Floodlight Eliote

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

---

Frigate+ would significantly help with your situation. The issue you're experiencing—dogs being misclassified as people due to unusual camera angles—is precisely what Frigate+ is designed to solve.(1)

Frigate+ provides community-trained models built from real security camera footage, which delivers far better accuracy than generic models.(1) More importantly, Frigate+ includes a fine-tuning workflow where you can upload images from your specific camera angles and correct false positives.(1) This allows you to create scene-specific models adapted to your high-mounted cameras.(1)

Since your cameras are mounted high with atypical viewing angles, the default model (trained on more standard perspectives) struggles to distinguish dogs from people in these conditions.(1) Frigate+ models are trained on a proprietary dataset sourced from real Frigate installations worldwide, providing accuracy beyond generic COCO-based models.(1)

What Frigate+ offers for your use case:

• Models trained on millions of real security camera images from various angles and conditions(1)
• A fine-tuning workflow to upload your own images and correct misclassifications(1)
• Expanded object vocabulary with better differentiation between similar objects(1)
• Continuous improvement as more community data is contributed(1)

The fact that Frigate's AI summary correctly identifies "two dogs on a wooden deck" suggests the underlying image understanding is there—Frigate+ would bring that level of accuracy to the real-time object detection.

For configuration adjustments without Frigate+, you can set per-object type filters with different thresholds, but this won't solve the fundamental misclassification issue.(2) You could lower the threshold for dogs specifically or adjust min_score, but as you noted, this may create other detection issues.(2)

Frigate+ is your best solution here since it addresses the root cause: training data that doesn't match your camera's perspective.(1)

📚 Sources:

1. Supported Hardware
2. Full Reference Config

---

Have another question? Just tag @inkeep.