openface/docs/usage.md

Usage

API Documentation

Example

See the image comparison demo for a complete example written in Python using a naive Torch subprocess to process the faces.

import openface

# `args` are parsed command-line arguments.

align = openface.AlignDlib(args.dlibFacePredictor)
net = openface.TorchNeuralNet(args.networkModel, args.imgDim, cuda=args.cuda)

# `img` is a numpy matrix containing the RGB pixels of the image.
bb = align.getLargestFaceBoundingBox(img)
alignedFace = align.align_v1(args.imgDim, img, bb,
                             landmarkIndices=openface.AlignDlib.OUTER_EYES_AND_NOSE)
rep1 = net.forward(alignedFace)

# `rep2` obtained similarly.
d = rep1 - rep2
distance = np.dot(d, d)