hi to everyone, sorry for being dumb perhaps but whatever I try I do not succeed in understanding what is explained in Reducing false positives | Frigate ie: line 3 says that with the values 0.7, 0.0 and 0.85 the computed score is 0.7 but (0.7+0.0+0.85)/3=0.516 and not 0.7… in the same way the following line (0.7+0.85+0.95+0.9)/4=0.85 and not 0.875. I’m confused there must be something I’m missing. Moreover whenever frigate reacts to something it gives two numbers on top of the box, I suppose the second one relates to the areas involved thus allowing to set a sort of filter, what I don’t understand is how to relate the first value (%) to the score and threshold settings. Lets say I have a false positive showing 59% what do I have to change in order to discard everything below 60%. I’m using frigate now since quite a long time and it’s performance is quite amazing but now and again I get those false positive which are very unwelcome especially at night, so I’ll be grateful to anyone willing to help. thank you
That is the average (otherwise known as mean), meanwhile the docs state that it uses the median. Meaning when the numbers are sorted [0.0, 0.7, 0.85] then the middle score is used.
min_score is the score needed to start tracking something (but won’t create an event). Threshold is the score that is needed to start tracking. The % shown in the snapshot is the current median score, but do keep in mind once an objects median is above the threshold it will continue to be tracked, so you need to consider that. Maybe start by increasing the threshold to 0.75 or so and see how that fairs.
Many thanks for spending your time to help me out, I will try to follow your suggestions nevertheless I need some clarifications: am I right to think that the “median value” among odd amount of figures is the center one while among even amount is the average between the “center ones”? Am I right to assume that min_score and threshold both start tracking while the only one which creates an event is threshold? Anyway as you said the only thing I may try to do is slowly increase the threshold. Again thank you very much.
Yes that is correct
I am not sure I follow what you mean there. min_score is what is required to start tracking, if the score was above threshold then naturally it would also be above min_score. However, an event won’t be created until the score has crossed the threshold
ok thanks again, it seems to be completely clear I’ll try to work on that 