'''

This is a file that will train a ML network given a set of images.

'''

import os

from os import listdir

from os.path import isfile, join

import skimage

from skimage import io, color

from skimage.feature import hog

from sklearn import svm

from sklearn.externals import joblib

from tqdm import tqdm

size = (400, 400)

ImagesPath = "./python/trainingImages"

class SVM:

svm = None

instantiated = False

def __init__(self):

self.load()

print "Instantiating the SVM"

if self.svm == None:

self.svm = svm.LinearSVC(verbose=4)

self.instantiated = True

def loadImages(self, test=False):

avPath = "./python/trainingImages/AverageLooking"

avImages = [f for f in listdir(avPath) if isfile(join(avPath, f))]

gdPath = "./python/trainingImages/GoodLooking"

gdImages = [f for f in listdir(gdPath) if isfile(join(gdPath, f))]

if test:

avImages, gdImages = avImages[:10], gdImages[:10]

averageImages = []

goodImages = []

for img in tqdm(avImages, "Import Average Photos"):

try:

averageImages.append(skimage.transform.resize(color.rgb2grey(io.imread(avPath + "/" + img)), size))

except:

os.remove(avPath + "/" + img)

continue

for img in tqdm(gdImages, "Import Good Photos"):

try:

goodImages.append(skimage.transform.resize(color.rgb2grey(io.imread(gdPath + "/" + img)), size))

except:

os.remove(gdPath + "/" + img)

continue

return goodImages, averageImages

def getFeatures(self, goodImages, avImages):

data = []

vals = []

for good, av in tqdm(zip(goodImages, avImages), desc="Extract features"):

try:

fdGood, hog_image = hog(good, orientations=8, pixels_per_cell=(16, 16), cells_per_block=(1, 1),

visualise=True)

fdAv, hog_image = hog(av, orientations=8, pixels_per_cell=(16, 16), cells_per_block=(1, 1), visualise=True)

except:

continue

data.append(fdGood)

vals.append("Good")

data.append(fdAv)

vals.append("Average")

print "Done extracting features"

return data, vals

def train(self, data, values):

'''

This function will train the SVM network on it's data to learn to understand the expected values

:param data:

:return:

'''

self.svm.fit(data, values)

print "Done training"

def save(self):

joblib.dump(self.svm, "./python/SVM_DATA.dat")

def load(self):

self.svm = joblib.load("./python/SVM_DATA.dat")

def predict(self, data):

'''

This function will predict the SVM values of a whole lot of images

:param data:

:return:

'''

data = skimage.transform.resize(color.rgb2grey(data), size)

fd, hog_image = hog(data, orientations=8, pixels_per_cell=(16, 16), cells_per_block=(1, 1), visualise=True)

return self.svm.predict(fd), self.svm.decision_function(fd)

def rate(img):

return [True, 0.94]

s = SVM()

prediction, decision = s.predict(img)

def main():

# s = SVM()

# goodImgs, avImgs = s.loadImages(test=False)

# imageData, imageVals = s.getFeatures(goodImgs, avImgs)

# s.train(imageData, imageVals)

# s.save()

s = SVM()

me = io.imread("/Users/eknox/workspace/InterestRate/python/trainingImages/GoodLooking/1.jpg")

prediction, decision = s.predict(me)

print "We are "+ str(100*decision[0])+"% sure that Evan is : " + str(prediction[0])

if __name__ == '__main__':

main()