def main():
print("PCA")
pca.pca()
print("RandomForest")
rf.rf(2)
print("KNN")
knn.knn(2)
print("SVC")
svc.svc()
print("GRID_SVC")
svc.gridSearchScore()
print("Logistic")
logistic.Logistic().fit()
print("DNN Classifier")
classifier_model = classifier.classifier()
classifier_model.fit()
# load smile data
PATH = "../data/smile/"
for idx, filename in enumerate(smilefiles):
phi[idx] = vectorize(PATH + filename)
labels.append(1)
# load neutral data
PATH = "../data/neutral/"
offset = idx + 1
for idx, filename in enumerate(neutralfiles):
phi[idx + offset] = vectorize(PATH + filename)
labels.append(0)
"""
training the data with logistic regression
"""
lr = logistic.Logistic(dim)
lr.train(phi, labels)
"""
open webcam and capture images
"""
cv2.namedWindow("preview")
vc = cv2.VideoCapture(0)
if vc.isOpened(): # try to get the first frame
rval, frame = vc.read()
else:
rval = False
print "\n\n\n\n\npress space to take picture; press ESC to exit"
while True:
def getTrainingRes():
"""
given a jpg image, vectorize the grayscale pixels to
a (width * height, 1) np array
it is used to preprocess the data and transform it to feature space
"""
def vectorize(filename):
size = defaults.WIDTH, defaults.HEIGHT # (width, height)
im = Image.open(filename)
resized_im = im.resize(size, Image.ANTIALIAS) # resize image
im_grey = resized_im.convert('L') # convert the image to *greyscale*
im_array = np.array(im_grey) # convert to np array
oned_array = im_array.reshape(1, size[0] * size[1])
return oned_array
"""
load training data
"""
# create a list for filenames of smiles pictures
smilefiles = []
with open(defaults.smile_csv, 'rb') as csvfile:
for rec in csv.reader(csvfile, delimiter=' '):
smilefiles += rec
# create a list for filenames of neutral pictures
neutralfiles = []
with open(defaults.neutral_csv, 'rb') as csvfile:
for rec in csv.reader(csvfile, delimiter=' '):
neutralfiles += rec
# create a list for filenames of disgust pictures
disgustfiles = []
with open(defaults.disgust_csv, 'rb') as csvfile:
for rec in csv.reader(csvfile, delimiter=' '):
disgustfiles += rec
# N x dim matrix to store the vectorized data (aka feature space)
phi = np.zeros((len(smilefiles) + len(neutralfiles) + len(disgustfiles),
defaults.dim))
# 1 x N vector to store binary labels of the data: 1 for smile and 0 for neutral
labels = []
# load smile data
for idx, filename in enumerate(smilefiles):
phi[idx] = vectorize(defaults.smile_imgs + filename)
labels.append(0)
# load neutral data
offset = idx + 1
for idx, filename in enumerate(neutralfiles):
phi[idx + offset] = vectorize(defaults.neutral_imgs + filename)
labels.append(1)
# load disgust data
offset = idx + 1
for idx, filename in enumerate(disgustfiles):
phi[idx + offset] = vectorize(defaults.disgust_imgs + filename)
labels.append(2)
"""
training the data with logistic regression
"""
lr = logistic.Logistic(defaults.dim)
lr.train(phi, labels)
return lr
#Convert the picture into a numpy array
buff = np.fromstring(stream.getvalue(), dtype=np.uint8)
#Now creates an OpenCV image
image = cv2.imdecode(buff, 1)
return image
if __name__ == '__main__':
## Exists weigths
if sys.argv[1]=="train":
#Load training data
lr=getTrainingRes()
else:
weigths = np.load(defaults.weigths_file)
lr = logistic.Logistic(defaults.dim,weigths)
"""
open webcam and capture images
"""
if defaults.SHOW_PREVIEW:
cv2.namedWindow("preview")
print "\n--Starting to get images"
if defaults.raspberry_plt:
cam = picamera.PiCamera()
cam.resolution = (320, 240)
else:
cam = cv.CaptureFromCAM(0)
while True:
if defaults.raspberry_plt: