[Support]: False positive vs poorly framed objects as they pertain to Frigate+ labeling

[nickcmeyer]

Describe the problem you are having

What are the limits to false vs true positives for poorly framed, overlapping, and blurry objects? As you can see in the background there are multiple cars staggered. I'm wondering if labeling these cars (including the full extent of the car especially when partially hidden behind another car, marked as difficult) has resulted in more partial vehicles being detected even when the full vehicle is visible? Basically thinking that labeling these may have taught the model to offset a bounding box based on the foreground vehicle more than center a bounding box on the background vehicle even when the background vehicle is clearly identifiable by a human.

Now that I have these partial detections should I be marking them as false positives or true positive and re-framing? That likely includes relabeling all of the staggered vehicles in the back at the same time, again. Or going back and trying to delete the vehicles not in the foreground from past images? Kind of wish I could just mask out the far reaches when sending to frigate+ so that I don't have to worry about certain things polluting the model.

Version

14

What browser(s) are you using?

No response

Frigate config file

NA

Relevant Frigate log output

NA

Relevant go2rtc log output

NA

FFprobe output from your camera

NA

Frigate stats

No response

Install method

Docker Compose

docker-compose file or Docker CLI command

NA

Object Detector

Coral

Network connection

Wired

Camera make and model

NA

Screenshots of the Frigate UI's System metrics pages

No response

Any other information that may be helpful

No response

[NickM-27]

In this case it is fine to mark as true positive and redo the box. False positive is mainly used when there is no true object in the same area.

And it's important that all objects are labeled. I have some similar views on my panoramic wide angle camera and haven't had any issues like this.