I am working on a simple face detection application. The workflow is pretty simple. If an image containing a person was taken with an iPhone camera then I infer face data using the xmp/mwg metadata. Else, I use OpenIMAJ's HaarCascadeDetector.
For the same image, the reported confidence values are different. For example, the value of the metadata attribute apple-fi:ConfidenceLevel is 283, whilst the value from the corresponding DetectedFace object is 47.
Clearly the two values are being calculated differently... only I can't see if/where Apple have published their methodology.
Does anyone know how I could approximately transform one value to the other?
(Apologies in advance if this question had already been answered, or indeed is too naive.)
Cheers,
Anthony
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
As far as I'm aware Apple don't publish any info on their implementation. The OpenIMAJ implementation of confidence for the HaarCascadeDetector is unnormalised and is based on the number of independent detections from the grouping stage of the cascade process.
It might be possible to figure out some kind of approximate relationship between the confidences of the two different implementations, however you might find they are almost completely uncorrelated.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
View and moderate all "General Discussion" comments posted by this user
Mark all as spam, and block user from posting to "Discussion"
Hi,
I am working on a simple face detection application. The workflow is pretty simple. If an image containing a person was taken with an iPhone camera then I infer face data using the xmp/mwg metadata. Else, I use OpenIMAJ's HaarCascadeDetector.
For the same image, the reported confidence values are different. For example, the value of the metadata attribute apple-fi:ConfidenceLevel is 283, whilst the value from the corresponding DetectedFace object is 47.
Clearly the two values are being calculated differently... only I can't see if/where Apple have published their methodology.
Does anyone know how I could approximately transform one value to the other?
(Apologies in advance if this question had already been answered, or indeed is too naive.)
Cheers,
Anthony
As far as I'm aware Apple don't publish any info on their implementation. The OpenIMAJ implementation of confidence for the HaarCascadeDetector is unnormalised and is based on the number of independent detections from the grouping stage of the cascade process.
It might be possible to figure out some kind of approximate relationship between the confidences of the two different implementations, however you might find they are almost completely uncorrelated.