[Support]: Should I annotate object I don't want to be detected in a particular camera

[0xskynet]

Describe the problem you are having

I am subscribed to frigate+ and annotating images. I have several cameras like driveway, porch, backyard etc. In the Frigate+ portal settings I have selected person and package to be detected on the porch camera. I don't care about detection like car. Do I need to annotate car in the porch camera images even if I don't want the cars to be detected in porch camera?

If yes, it does not allow me to select car in the annotation because in "Annotation Labels" I haven't selected "Car" for the porch edit camera option.
I read the documentation but I am confused by,

Label every object in the image: It is important that you label all objects in each image before verifying. If you don't label a car for example, the model will be taught that part of the image is not a car and it will start to get confused. You can exclude labels that you don't want detected on any of your cameras.

You can exclude labels that you don't want detected on any of your cameras

Thanks!

Version

0.17.1-416a9b7

What browser(s) are you using?

Chrome

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

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

Yes, you should annotate all objects (including cars) in your porch camera images, even if you don't want cars detected on that specific camera. (1)

The key distinction in the documentation is:

• "Label every object in the image" – This applies to individual images you're annotating. If a car appears in a porch camera image, you should label it. Otherwise, the model learns that region is "not a car," which causes confusion. (1)

• "You can exclude labels that you don't want detected on any of your cameras" – This refers to labels you want excluded across your entire Frigate deployment, not per-camera exclusions. (1)

For your situation: If you want cars detected on your driveway camera but not your porch camera, you should still enable the "car" label in your Frigate+ annotation settings for all cameras (including porch) so you can properly annotate cars when they appear. The per-camera detection filtering should be handled in your Frigate NVR configuration using the objects.track setting, not by excluding labels from annotation. (2)

In your Frigate config, you can configure each camera to only track specific objects:

cameras:
  porch:
    objects:
      track:
        - person
        - package
  driveway:
    objects:
      track:
        - person
        - car

(2)

This way your model learns what cars look like (improving overall accuracy), but your porch camera only triggers on person and package detections.

📚 Sources:

1. Models
2. Full Reference Config

---

Have another question? Just tag @inkeep.