Closed deepfree2023 closed 4 months ago
it's not an error, the scores are not absolute but relative to the model. We could do normalization I guess not sure what is the right strategy for that
I see. Is it true that different thresholds are required for different face analysis models?
yes. that is true. I will add deepface next and see if there's a way to normalize the results
The scores obtained by insightface are much larger than the ones from dlib. Could this be fixed?
Dib:
Insightface: