Constructor

obj = cv.Facemark(ftype)
obj = cv.Facemark(ftype, 'OptionName',optionValue, ...)

Input

• ftype Facemark algorithm, one of:
  • LBF regressed local binary features (LBF).
  • AAM active appearance model (AAM).

Options for LBF

• ShapeOffset offset for the loaded face landmark points. default 0.0
• CascadeFace filename of the face detector model. default ''
• Verbose show the training print-out. default true
• NLandmarks number of landmark points. default 68
• InitShapeN multiplier for augment the training data. default 10
• StagesN number of refinement stages. default 5
• TreeN number of tree in the model for each landmark point refinement. default 6
• TreeDepth the depth of decision tree, defines the size of feature. default 5
• BaggingOverlap overlap ratio for training the LBF feature. default 0.4
• ModelFilename filename where the trained model will be saved (Base64 encoded). default ''
• SaveModel flag to save the trained model or not. default true
• Seed seed for shuffling the training data. default 0
• FeatsM default [500,500,500,300,300,300,200,200,200,100]
• RadiusM default [0.3,0.2,0.15,0.12,0.10,0.10,0.08,0.06,0.06,0.05]
• Pupils index of facemark points on pupils of left and right eye. default {[36,37,38,39,40,41], [42,43,44,45,46,47]}
• DetectROI default [-1,-1,-1,-1]

Options for AAM

• ModelFilename filename where the trained model will be saved (Base64 encoded). default ''
• SaveModel flag to save the trained model or not. default true
• M default 200
• N default 10
• NIter default 50
• Verbose show the training print-out. default true
• MaxM default 550
• MaxN default 136
• TextureMaxM default 145
• Scales the scales considered to build the model. default [1.0,]