Closed Thedougler closed 11 months ago
It's a valid question, and it's a minimum fraction for tags
, it is not related to the value or weights.
In a batch interrogation, if you have 10 images and 3 images gave black hair as tag, above other thresholds, then setting this minimum fraction
to 0.2 will retain black hair whereas a fraction of 0.33 won't (because 3 in 10 is 0.30 which is less than 0.33). Note that you can combine interrogations (checkbox) and then this minimum fraction
will filter across all the interrogations of the images in your batch,
The minimum fraction
can also apply for a single image, with these combined interrogations; if you used 5 different kinds of interrogators, and two yielded brown hair that's a fraction of 0.4.
At least that's how this is supposed to work. In the future I may allow to give each interrogation, in combined interrogations a weight, and then this will make the fraction a bit more difficult to explain, though the principles should stay the same. Hope that makes sense.
Ahh so it's like a "Percent required for positive results"
Yes, My rationale for adding this slider was: oftentimes people use this extension to train for particular features and have collected images with these features, that consequently are very common in their batch of images. This slider makes it easier to select exactly those features that are common in the set. I renamed the slider label, I hope that makes it more clear.
So maybe im being dense but its not clear what function the "Minimum Value for Tags: field does