Open topic with navigation
Bayesian classifiers, and the
Priors configuration parameter, are deprecated in Media Server 11.4.0 and later. HPE recommends that you use Convolutional Neural Network (CNN) classifiers instead.
This parameter is still available for existing implementations, but it might be incompatible with new functionality. The parameter might be removed in future.
The probability that object classes will be seen in the media being analyzed. If you do not set this parameter, the image classification task expects that all of the classes are equally likely to be present in the media.
This parameter accepts a list of object class names mapped to floating point numbers. The numbers must be proportional to the prior probability that the object class will be seen. You must assign a value to every class that exists in the classifier.
You can only use this parameter with Bayesian classifiers. The type of classifier is specified with the
ClassifierType training option (see SetClassifierTrainingOption).
|Type:||String (comma-separated list)|
In the following example, the image classification task expects that more images or video frames will be classified as images of cars than other types of vehicle:
Priors0=cars : 5 Priors1=vans : 3 Priors2=hgvs : 2