def check_distances(self):
for i in range(self.node_idx):
distance = dcos(self.G.node[i]['desc'],
self.G.node[self.node_idx]['desc'])
# distance = abs(self.G.node[i]['desc'] - self.G.node[self.node_idx]['desc'])
if distance <= self.threshold:
self.G.add_edge(i, self.node_idx, weight=distance)
def find_closest(img, database, min_detection=2.5):
imarr1 = np.asarray(img)
imarr1 = imarr1[None, ...]
#Prediction
fvec1 = featuremodel.predict(imarr1)[0, :]
#Closest person in DB
dmin = 0.0
umin = ""
for key, value in database.items():
fvec2 = value
dcos_1_2 = dcos(fvec1, fvec2)
if umin == "":
dmin = dcos_1_2
umin = key
elif dcos_1_2 < dmin:
dmin = dcos_1_2
umin = key
if dmin > min_detection:
umin = ""
return umin, dmin
def calculate_distance(im1, im2):
fvec1 = extract_feature(im1)
fvec2 = extract_feature(im2)
dcos_1_2 = dcos(fvec1, fvec2)
print('----------: {}'.format(dcos_1_2))
return dcos_1_2
def cos_distance(feature_vector1, feature_vector2):
return dcos(feature_vector1, feature_vector2)
def get_distance(self, node0, node1):
return dcos(self.G.node[node0]['desc'], self.G.node[node1]['desc'])
# convert the image pixels to a numpy array
imarr_i = img_to_array(imarr_i)
# reshape data for the model
imarr_i = imarr_i.reshape(
(1, imarr_i.shape[0], imarr_i.shape[1], imarr_i.shape[2]))
return imarr_i
imarr1 = get_im(str1)
imarr2 = get_im(str2)
imarr3 = get_im(str3)
#imarr3=get_im(str3)
#imarr4=get_im(str4)
imarr_t = crpim
imarr_t = img_to_array(imarr_t)
# reshape data for the model
imarr_t = imarr_t.reshape(
(1, imarr_t.shape[0], imarr_t.shape[1], imarr_t.shape[2]))
#imarr2=Image.open(crpim)
face_id = {str1: imarr1, str2: imarr2, str3: imarr3}
featuremodel = Model(input=facemodel.layers[0].input,
output=facemodel.layers[-2].output)
from scipy.spatial.distance import cosine as dcos
for j in face_id:
fvec1 = featuremodel.predict(face_id[j])[0, :]
fvec2 = featuremodel.predict(imarr_t)[0, :]
dcos_1_2 = dcos(fvec1, fvec2)
if (dcos_1_2 < 0.2):
print('\n \n \n RECOGNISED {0}'.format(j))