def testing_run(frame):
try:
img = frame
#face_locations = face_recognition.face_locations(img)
face_locations = face_recognition.face_locations(
image, number_of_times_to_upsample=0, model="cnn")
#print("I found {} face(s) in this photograph.".format(len(face_locations)))
for face_location in face_locations:
# Print the location of each face in this image
top, right, bottom, left = face_location
#print("A face is located at pixel location Top: {}, Left: {}, Bottom: {}, Right: {}".format(top, left, bottom, right))
# You can access the actual face itself like this:
face_image = img[top:bottom, left:right]
pil_image = Image.fromarray(face_image)
example_idx = 20
example_idx = 20
example_image = np.asarray(pil_image)
#example_image =cv2.cvtColor(example_image, cv2.COLOR_BGR2GRAY)
# obtain embedding vector for image
myemd = return_emd(example_image)
example_prediction = svc.predict([myemd])
#imf = cv2.cvtColor(face_recognition.load_image_file("myimg.jpg"), cv2.COLOR_BGR2GRAY)
fd = face1.face_recognition.face_encodings(example_image)[0]
example_prediction1 = svc1.predict([fd])
#print("svc1 pred: "+str(np.amax(svc1.predict_proba([fd]))))
example_identity = encoder.inverse_transform(example_prediction)[0]
example_identity1 = encoder1.inverse_transform(
example_prediction1)[0]
#print(str(example_identity)+" -- "+str(example_identity1))
thresh = 0.25
#print(face1.distance(mic[(example_identity1,0)],fd))
#print(face1.distance(mic[(example_identity1,1)],fd))
if face1.distance(mic[(example_identity1, 0)],
fd) < thresh and face1.distance(
mic[(example_identity1, 1)], fd) < thresh:
print(example_identity1)
return example_identity1
else:
print("unknown")
return "unknown"
#plt.imshow(example_image)
#plt.title(f'Recognized as {example_identity}');
except Exception as e:
print(e)
def testing_run(frame):
global thresh
if 1:
img = frame
face_locations = face_recognition.face_locations(img)
#print("I found {} face(s) in this photograph.".format(len(face_locations)))
if len(face_locations) == 0:
print("noface")
#data_maintain("noface",img,np.array(np.random(1,128)))
else:
for face_location in face_locations:
# Print the location of each face in this image
top, right, bottom, left = face_location
# You can access the actual face itself like this:
face_image = img[top:bottom, left:right]
#cv2.rectangle(frame, (left-5, top-5), (right-5, bottom+5), (255,0,0), 2)
cv2.imwrite('sada.jpg', face_image)
cv2.waitKey(0)
example_image = np.asarray(face_image)
fd = face_recognition.face_encodings(example_image)[0]
example_prediction1 = knn1.predict([fd])
example_identity1 = encoder1.inverse_transform(
example_prediction1)[0]
if face1.distance(mic[
(example_identity1, 1)], fd) < thresh and face1.distance(
mic[(example_identity1, 2)], fd) < thresh:
print(example_identity1)
# data_maintain(example_identity1, face_image,fd)
# elif face1.distance(mic[(example_identity1,0)],fd) < thresh2 and face1.distance(mic[(example_identity1,1)],fd) < thresh2 :
# print(example_identity1)
# print("2")
# for_asking(example_identity1, face_image,fd)
else:
print("unknown")
def retrain():
now = datetime.datetime.now()
global past_time
if past_time == None:
past_time = now
time_diff =int((now-past_time).total_seconds())/60
import platform
if platform.system() == 'Windows':
win =1
else:
win =0
if 1:#time_diff >1:
global main_c
past_time = now
# while global_lock.locked():
# continue
# global_lock.acquire()
print('New Face adding...')
main_c.execute("select * from add_person")
data = main_c.fetchall()
main_c.execute("truncate table add_person")
f = h5py.File('tempFiles/temp.hdf5','a')
g = h5py.File('tempFiles/embedded2.hdf5','a')
for d in data:
if 1:
iemd = f[d[1]].value
del f[d[1]]
gg = g.create_group(d[2])
print("XXX")
if win !=1:
os.system("mkdir images/"+str(d[2]))
os.system("mv temp_image/"+str(d[1])+" images/"+str(d[2])+"/1.jpg")
print("sadas")
else:
os.system("md images\\"+str(d[2]))
os.system("move temp_image\\"+str(d[1])+" images\\"+str(d[2])+"\\1.jpg")
no_img=1
gg.create_dataset(str(no_img)+".jpg", data=iemd)
for key in f.keys():
if face1.distance(f[key].value,iemd) < 0.25:
print("in keys")
main_c.execute('delete from notification where img_name=%s',(key))
no_img=no_img+1
if win !=1:
os.system("mv temp_image/"+str(key)+" images/"+str(d[2])+"/"+str(no_img)+".jpg")
else:
os.system("move temp_image\\"+str(key)+" images\\"+str(d[2])+"\\"+str(no_img)+".jpg")
gg.create_dataset(str(no_img)+".jpg", data=f[key].value)
del f[key]
while no_img<3:
print(no_img)
nn = no_img+1
if win !=1:
os.system("cp images/"+str(d[2])+"/"+str(no_img)+".jpg images/"+str(d[2])+"/"+str(nn)+".jpg")
else:
os.system("copy images\\"+str(d[2])+"\\"+str(no_img)+".jpg images\\"+str(d[2])+"\\"+str(nn)+".jpg")
no_img = no_img+1
gg.create_dataset(str(no_img)+".jpg", data=iemd)
main_c.execute('insert into user_detail(name,no_img,last_update_date) values(%s,%s,%s)',(d[2],no_img,now.strftime("%Y-%m-%d")))
db.commit()
# except Exception as e:
# print(e)
# main_c.execute("select * from for_asking where flag = 1")
# data1 = main_c.fetchall()
# main_c.execute("delete from for_asking where flag=1")
# for d1 in data1:
# local_c.execute("select no_img from user_detail where name = %s",(d1[1]))
# no_img = local_c.fetchone()
# no = no_img[0] + 1
# sf = len(os.listdir(os.pajoin(path, d1[1])))
# s =int(sf)+1
# local_c.execute("update user_detail set no_img = %s",(no))
# os.system("mv temp_image/"+str(d1[2])+" images/"+str(d1[1])+"/"+s+".jpg")
f.close()
g.close()
#global_lock.release()
print('### complete face adding ###') #retrain model
if not len(data) == 0:
pre_load()
def pre_load():
global knn1, svc1, encoder1, mic
metadata = face1.load_metadata('images')
num = len(metadata)
filename = 'tempFiles/embedded2.hdf5'
f = h5py.File(filename, 'r')
for raw in f:
for j in f[raw]:
xc = j.split('.')
mic[str(raw), int(xc[0])] = f[raw][j][:]
embedded2 = np.zeros((metadata.shape[0], 128))
for i, m in enumerate(metadata):
embedded2[i] = f[m.name + "/" + m.file].value
f.close()
train_in = np.ones((metadata.shape[0]), dtype=bool)
test_in = np.ones((metadata.shape[0]), dtype=bool)
name = None
for i, m in enumerate(metadata): # 90 / 10 dataset
if name == None:
name = m.name
train_in[i] = False
test_in[i] = True
continue
if name == m.name:
train_in[i] = True
test_in[i] = False
else:
name = m.name
train_in[i] = False
test_in[i] = True
distances = [] # squared L2 distance between pairs
identical = [] # 1 if same identity, 0 otherwise
num = len(metadata)
for i in range(num - 1):
for j in range(1, num):
distances.append(face1.distance(embedded2[i], embedded2[j]))
identical.append(1 if metadata[i].name == metadata[j].name else 0)
distances = np.array(distances)
identical = np.array(identical)
thresholds = np.arange(0.2, 0.4, 0.005)
#print(thresholds)
# for t in thresholds:
# print(identical)
# print(f1_score(identical, distances < t))
f1_scores = [f1_score(identical, distances < t) for t in thresholds]
acc_scores = [accuracy_score(identical, distances < t) for t in thresholds]
opt_idx = np.argmax(f1_scores)
# Threshold at maximal F1 score
opt_tau = thresholds[opt_idx]
# Accuracy at maximal F1 score
opt_acc = accuracy_score(identical, distances < opt_tau)
global thresh
thresh = opt_tau + 0.02
print(f'Accuracy at threshold {opt_tau:.3f} = {opt_acc:.3f}')
#####################################
targets2 = np.array([m.name for m in metadata])
encoder1 = LabelEncoder()
encoder1.fit(targets2)
y = encoder1.transform(targets2)
# 50 train examples of 10 identities (5 examples each)
X_train = embedded2[train_in]
# 50 test examples of 10 identities (5 examples each)
X_test = embedded2[test_in]
y_train = y[train_in]
y_test = y[test_in]
knn1.fit(X_train, y_train)
svc1.fit(X_train, y_train)
acc_knn1 = accuracy_score(y_test, knn1.predict(X_test))
acc_svc1 = accuracy_score(y_test, svc1.predict(X_test))
print(f'KNN accuracy = {acc_knn1}, SVM accuracy = {acc_svc1}')
def testing_run(frame):
#try:
img = frame
face_locations = face_recognition.face_locations(img)
#print("I found {} face(s) in this photograph.".format(len(face_locations)))
if len(face_locations) == 0:
print("noface")
#data_maintain("noface",img,np.array(np.random(1,128)))
else:
cv2.imwrite("a.jpg", img)
unknown_image = face_recognition.load_image_file("a.jpg")
pp = Image.fromarray(unknown_image)
draw = ImageDraw.Draw(pp)
a = []
for face_location in face_locations:
# Print the location of each face in this image
top, right, bottom, left = face_location
# You can access the actual face itself like this:
face_image = img[top:bottom, left:right]
pil_image = Image.fromarray(face_image)
draw.rectangle(((left - 5, top - 5), (right - 5, bottom + 5)),
outline=(200, 0, 0))
example_image = np.asarray(face_image)
fd = face1.face_recognition.face_encodings(example_image)[0]
example_prediction1 = svc1.predict([fd])
example_identity1 = encoder1.inverse_transform(
example_prediction1)[0]
# Draw a label with a name below the face
thresh = 0.30
thresh2 = 0.40
if face1.distance(mic[(example_identity1, 0)],
fd) < thresh and face1.distance(
mic[(example_identity1, 1)], fd) < thresh:
print(example_identity1)
text_width, text_height = draw.textsize(str(example_identity1))
draw.rectangle(((left - 5, bottom - text_height - 10 - 5),
(right - 5, bottom + 5)),
fill=(200, 0, 0),
outline=(200, 0, 0))
draw.text((left + 6, bottom - text_height - 5),
str(example_identity1),
fill=(255, 255, 255, 255))
a.append(example_identity1)
# elif face1.distance(mic[(example_identity1,0)],fd) < thresh2 and face1.distance(mic[(example_identity1,1)],fd) < thresh2 :
# print(example_identity1)
# print("2")
# for_asking(example_identity1, face_image,fd)
else:
print("unknown")
a.append("unknown")
text_width, text_height = draw.textsize("Unknown")
draw.rectangle(
((left, bottom - text_height - 10), (right, bottom)),
fill=(0, 0, 255),
outline=(0, 0, 255))
draw.text((left + 6, bottom - text_height - 5),
"Unknown",
fill=(255, 255, 255, 255))
del draw
dx = PyMySQL.connect("localhost", "root", "", "attendence")
cr = dx.cursor()
now = datetime.datetime.now()
cr.execute("INSERT INTO data(datetim) values(%s)",
(now.strftime('%H:%M %Y-%m-%d')))
dx.commit()
for x in a:
aq = ("UPDATE data SET {}=1 where datetim='{}'").format(
x, now.strftime('%H:%M %Y-%m-%d'))
cr.execute(aq)
dx.commit()
cr.close()
dx.close()
pp.show()