class FaceRecognitionModel():
# List of people
def __init__(self):
self.SVM = Classifier(verbose=True)
self.processor = Processor(verbose=True)
self.util = util()
def learn(self, dir):
"""
Learns a new face from a set of images in a directory.
"""
imgs = openface.data.iterImgs(dir)
for imgObject in imgs:
self.processor.processImage(imgObject, isTrain=True)
# Train face detection model
self.SVM.train()
def recognize(self, img):
"""
Finds faces in an image and labels if they are recognized or not.
"""
# Create image object from img
imgName = os.path.basename(img)
imgObject = openface.data.Image("?", imgName, img)
# Get face rAepresentation: list of (bb, face vectors)
reps = self.processor.processImage(imgObject)
# Pass reps into SVM classifier to get predictions
faces = self.SVM.infer(
self.util.classifierModelDir,
reps) # (classifierModelDir, imgs,multiple=False,verbose=False)
print faces
# list of (person's name, confidence , bounding box)
# list of (string, float , rectangle object)
return faces
def find(self):
"""
Find all faces in the image and return locations (bounding boxes)
"""
return ""
def learnAll(self, dir):
"""
Learns from the entire dataset using the specified training-images directory.
"""
# process all images in the directory
imgs = openface.data.iterImgs(dir)
for imgObject in imgs:
reps = self.processor.processImage(imgObject, isTrain=True)
# Create a directory for generated embeddings (if not already created)
openface.helper.mkdirP(self.util.embeddingsDir)
# create CSV files using csvigo called from openface source code
subprocess.call([
'./batch-represent/main.lua', "-outDir", self.util.embeddingsDir,
"-data", self.util.alignedImgsDir
])
# TODO change string directory ^ for batch-represent to a define package
# Train face detection model
self.SVM.train()
# This will generate a new file called ./generated-embeddings/classifier.pkl.
# This file has the SVM model you'll use to recognize new face
print("Finished learning all knowledge!")
