WOLFRAM

Wolfram Language & System Documentation Center

Gender Classification

Train a classifier to recognize male from female faces in images.

Import a labeled face dataset.
Wolfram Language code: images = {[image], ..., [image]}; gender = {"male", ..., "female"};
Trim the faces out of the images and regularize the set using FindFaces, HistogramTransform, and ConformImages.
Wolfram Language code: faces = ConformImages[HistogramTransform@ImageTrim[#, FindFaces[#][[1]]]& /@ images, ColorSpace -> "Grayscale"]
Wolfram Language code: Out[1]= {[image], ..., [image]}
Define a training and a test set and prepare the classifier with the former.
Wolfram Language code: trainingSetSize = 30; trainingSet = Rule@@@Transpose@{faces, gender}[[All, ;; trainingSetSize]]; testSet = Rule@@@Transpose@{faces, gender}[[All, trainingSetSize + 1 ;; ]];
Use Classify to build the classifier.
Wolfram Language code: GenderClassifier = Classify[trainingSet, Method -> "LogisticRegression"]
Measure the accuracy of the classifier on the two sets.
Wolfram Language code: ClassifierMeasurements[GenderClassifier, trainingSet, "Accuracy"]
Wolfram Language code: ClassifierMeasurements[GenderClassifier, testSet, "Accuracy"]
Create a function that can classify arbitrary images.
Wolfram Language code: GenderDetect[img_Image] := With[{faces = FindFaces[img]}, If[faces =!= {}, GenderClassifier[ HistogramTransform[ImageTrim[ColorConvert[img, GrayLevel], faces[[1]]]] ], "no face detected" ]] GenderDetect[images__Image] := GenderDetect /@ {images}
Wolfram Language code: imgs = {[image], [image], [image], [image], [image]};
Wolfram Language code: GenderDetect@@imgs