Help reducing false positives

[AaronH]

I'm having generally good success with Sighthound, but would like to see if I can get the false positive rate down. I am actually wanting to detect when my dogs are outside (i.e. not people). So I use rules that look for unknown objects crossing boundaries. Ultimately it would be great if Sighthound could identify the dogs specifically, but until then maybe there are other workarounds? Here are the issues I'm seeing:

There are three situations that all have similar consequences: (a) When my camera adapts to various lighting conditions, it will go from B/W to Color or vice-versa. (b) At night when a light is turned on. (c) There is some occasional glitch in the stream. When any of these happen it often triggers an object detection. Most often Sighthound will think that the object is the entire frame, then the object shrinks quickly until it vanishes. This makes me think that a simple rule like "ignore object greater than" could be helpful. But I only see an "ignore object smaller than".

Thank you!

[cwizard]

Cameras changing from day to night trigger entire frame detection of a person for me too.... I'm also having an issue where a single strand of a spider web swaying in frame gets detected as a person.