def align_id_dataset(id_dataset, pnet, rnet, onet):
aligned_images = []
for i in range(len(id_dataset)):
image = misc.imread(os.path.expanduser(id_dataset[i].image_path), mode='RGB')
face_patches, _, _ = detect_and_align.align_image(image, pnet, rnet, onet)
aligned_images = aligned_images + face_patches
aligned_images = np.stack(aligned_images)
return aligned_images
def main(image_path, ready=False):
id_folder = ['ids']
test_folder = None
model = './model/'
with tf.Graph().as_default():
with tf.Session() as sess:
pnet, rnet, onet = detect_and_align.create_mtcnn(sess, None)
load_model(model)
images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0")
embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")
phase_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0")
id_dataset = id_data.get_id_data(id_folder[0], pnet, rnet, onet, sess, embeddings, images_placeholder, phase_train_placeholder)
print_id_dataset_table(id_dataset)
test_run(pnet, rnet, onet, sess, images_placeholder, phase_train_placeholder, embeddings, id_dataset, test_folder)
show_landmarks = False
show_bb = False
show_id = True
show_fps = False
if ready is False:
frame = cv2.imread(image_path)
elif ready is True:
frame = image_path
face_patches, padded_bounding_boxes, landmarks = detect_and_align.align_image(frame, pnet, rnet, onet)
if len(face_patches) > 0:
face_patches = np.stack(face_patches)
feed_dict = {images_placeholder: face_patches, phase_train_placeholder: False}
embs = sess.run(embeddings, feed_dict=feed_dict)
print('Matches in frame:')
for i in range(len(embs)):
bb = padded_bounding_boxes[i]
matching_id, dist = find_matching_id(id_dataset, embs[i, :])
if matching_id:
print('Hi %s! Distance: %1.4f' % (matching_id, dist))
else:
matching_id = 'Unknown'
print('Unkown! Couldn\'t fint match.')
return matching_id
else:
matching_id = 'Face Unfound'
print(matching_id)
return matching_id
def process_one(self):
if not self.to_process:
return
# input is an ascii string.
input_str = self.to_process.pop(0)
# convert it to a pil image
input_img = base64_to_pil_image(input_str)
################## where the hard work is done ############
# output_img is an PIL image
# output_img = input_img #self.makeup_artist.apply_makeup(input_img)
# output_str is a base64 string in ascii
# output_str = pil_image_to_base64(output_img)
# convert eh base64 string in ascii to base64 string in _bytes_
# self.to_output.append(binascii.a2b_base64(output_str))
open_cv_image = np.array(input_img)
# Convert RGB to BGR
open_cv_image = open_cv_image[:, :, ::-1].copy()
print("Processing frame...")
face_patches, padded_bounding_boxes, landmarks = detect_and_align.align_image(
open_cv_image, self.pnet, self.rnet, self.onet)
matching_id = "Unknown"
if len(face_patches) > 0:
face_patches = np.stack(face_patches)
feed_dict = {
self.images_placeholder: face_patches,
self.phase_train_placeholder: False
}
embs = self.sess.run(self.embeddings, feed_dict=feed_dict)
print('Matches in frame:')
for i in range(len(embs)):
bb = padded_bounding_boxes[i]
matching_id, dist = self.find_matching_id(embs[i, :])
if matching_id:
print('Hi %s! Distance: %1.4f' % (matching_id, dist))
else:
matching_id = 'Unknown'
print('Unknown! Couldn\'t fint match.')
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(open_cv_image, matching_id, (bb[0], bb[3]), font,
1, (255, 255, 255), 1, cv2.LINE_AA)
cv2.rectangle(open_cv_image, (bb[0], bb[1]), (bb[2], bb[3]),
(255, 0, 0), 2)
else:
print("No face patches")
match_dict = {}
match_dict[matching_id] = open_cv_image
# adding matching_name=>frame to array
self.to_output.append(match_dict)
def run():
with tf.Session() as sess:
LOG.log("Loading modell","SYSTEM")
#temp test
global pnet
global rnet
global onet
global images_placeholder
global embeddings
global phase_train_placeholder
pnet, rnet, onet = detect_and_align.create_mtcnn(sess, None)
model_exp = os.path.expanduser(model_path)
if (os.path.isfile(model_exp)):
# print('Model filename: %s' % model_exp)
with gfile.FastGFile(model_exp, 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
tf.import_graph_def(graph_def, name='')
images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0")
embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")
phase_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0")
# set up tensorflow model
#load_model(model_path)
LOG.log("Start system","SYSTEM")
while True:
if cam_cap.isOpened():
start = time.time()
# Get current frame
global frame
ret, frame = cam_cap.read()
# Count tick
global ticks
ticks = ticks + 1
# Do detection
global DETECTION_SLEEP_TICKS
if DETECTION_SLEEP_TICKS <= ticks:
#print("Detection")
global face_box
global face_found
global show_id
global show_bb
global show_landmarks
# Do detection
face_patches, padded_bounding_boxes, landmarks = detect_and_align.align_image(frame, pnet, rnet, onet)
# if found faces
if len(face_patches) > 0:
face_patches = np.stack(face_patches)
feed_dict = {images_placeholder: face_patches, phase_train_placeholder: False}
embs = sess.run(embeddings, feed_dict=feed_dict)
# print('Matches in frame:')
for i in range(len(embs)):
bb = padded_bounding_boxes[i]
if show_bb:
cv2.rectangle(frame, (bb[0], bb[1]), (bb[2], bb[3]), (255, 0, 0), 2)
if show_landmarks:
for j in range(5):
size = 1
top_left = (int(landmarks[i, j]) - size, int(landmarks[i, j + 5]) - size)
bottom_right = (int(landmarks[i, j]) + size, int(landmarks[i, j + 5]) + size)
cv2.rectangle(frame, top_left, bottom_right, (255, 0, 255), 2)
# Convert box to OpenCV
face_box = convert_tensorflow_box_to_openCV_box(padded_bounding_boxes[0])
# print (face_box)
# if running custom tracker this is needed
update_custom_tracker()
face_found = True
#return True
else:
# No face
face_found = False
#return False
# if face found
if face_found:
ticks = 0
global FAST_DETECTION_SLEEP_TICKS
DETECTION_SLEEP_TICKS = FAST_DETECTION_SLEEP_TICKS
else:
# Make less detections if not
ticks = 0
global SLOW_DETECTION_SLEEP_TICKS
DETECTION_SLEEP_TICKS = SLOW_DETECTION_SLEEP_TICKS
else:
# Do tracking
if face_found:
object_custom_tracking()
# print fps
end = time.time()
seconds = end - start
if seconds != 0:
fps = round(1 / seconds, 2)
if show_fps:
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(frame, str(fps), (0, 100), font, 1, (255, 255, 255), 1, cv2.LINE_AA)
#Show Cam
cv2.imshow('Detection GUI', frame)
#Close Program functionallity
if cv2.waitKey(25) & 0xFF == ord('q'):
cam_cap.release()
cv2.destroyAllWindows()
break
time.sleep(0.2) # Sleep
def main(args):
with tf.Graph().as_default():
with tf.Session() as sess:
faceCascade = cv2.CascadeClassifier(
'classifier/haarcascade_frontalface_default.xml')
file = os.path.join(os.getcwd(), "log.csv")
pnet, rnet, onet = detect_and_align.create_mtcnn(sess, None)
load_model(args.model)
images_placeholder = tf.get_default_graph().get_tensor_by_name(
"input:0")
embeddings = tf.get_default_graph().get_tensor_by_name(
"embeddings:0")
phase_train_placeholder = tf.get_default_graph(
).get_tensor_by_name("phase_train:0")
id_dataset = id_data.get_id_data(args.id_folder[0], pnet, rnet,
onet, sess, embeddings,
images_placeholder,
phase_train_placeholder)
print_id_dataset_table(id_dataset)
test_run(pnet, rnet, onet, sess, images_placeholder,
phase_train_placeholder, embeddings, id_dataset,
args.test_folder)
cap = cv2.VideoCapture(0)
cap.set(4, 9600)
cap.set(3, 12800)
frame_height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
int_frame_height = int(frame_height)
# img_path = os.path.join(args.id_folder)
# f = open("D:\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\log.csv", 'a')
f = open(
"C:\\Users\\Siva-Datta.Mannava\\OneDrive - Shell\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\log.csv",
'a')
# df = pd.read_csv("D:\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\log.csv")
df = pd.read_csv(
"C:\\Users\\Siva-Datta.Mannava\\OneDrive - Shell\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\log.csv"
)
# df = pd.read_csv(f)
# print("df", df)
# columns = pd.read_csv("D:\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\log.csv", nrows = 0)
# print('columns', columns)
count = 0
show_landmarks = False
show_bb = False
show_id = True
show_fps = False
total_people = 10
matched_ids = []
ids = os.listdir('ids')
while (True):
start = time.time()
_, frames = cap.read()
gray = cv2.cvtColor(frames, cv2.COLOR_BGR2GRAY)
count_msg = "Evacuated staff: " + str(count)
emergency_msg = "Emergency Evacuation"
remaining_msg = "Remaining staff: " + str(total_people - count)
cv2.putText(frames, emergency_msg, (0, 80),
cv2.FONT_HERSHEY_SIMPLEX, 3, (0, 0, 255), 3,
cv2.LINE_AA)
faces = faceCascade.detectMultiScale(
gray,
scaleFactor=1.05,
minNeighbors=8,
minSize=(55, 55),
flags=cv2.CASCADE_SCALE_IMAGE)
imgs = []
if len(ids) < 5:
show_missing(frames, ids, 0, int_frame_height)
for (x, y, w, h) in faces:
crop_img = frames[y - 50:y + h + 25, x - 25:x + w + 25]
imgs.append(crop_img)
for frame in imgs:
face_patches, padded_bounding_boxes, landmarks = detect_and_align.align_image(
frame, pnet, rnet, onet)
id_dataset = id_data.get_id_data(args.id_folder[0], pnet,
rnet, onet, sess,
embeddings,
images_placeholder,
phase_train_placeholder)
if len(face_patches) > 0:
face_patches = np.stack(face_patches)
feed_dict = {
images_placeholder: face_patches,
phase_train_placeholder: False
}
embs = sess.run(embeddings, feed_dict=feed_dict)
print('Matches in frame:')
for i in range(len(embs)):
bb = padded_bounding_boxes[i]
matching_id, dist = find_matching_id(
id_dataset, embs[i, :])
if matching_id:
print('Hi %s! Distance: %1.4f' %
(matching_id, dist))
if matching_id not in matched_ids:
matched_ids.append(matching_id)
count = count + 1
count_msg = "Evacuated staff: " + str(
count)
remaining_msg = "Remaining staff: " + str(
total_people - count)
if matching_id in ids:
ids.remove(matching_id)
else:
matching_id = 'Unkown'
print('Unkown! Couldn\'t fint match.')
'''Make folder and store the image'''
try:
# os.makedirs(os.path.join(args.id_folder, str(count)))
# dir1 = "D:\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\ids\\" + str(count)
dir1 = "C:\\Users\\Siva-Datta.Mannava\\OneDrive - Shell\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\ids\\" + str(
count)
os.makedirs(dir1)
except OSError as e:
if e.errno != errno.EEXIST:
raise
# path_img = "D:\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\ids\\" + str(count) + "\\"
path_img = "C:\\Users\\Siva-Datta.Mannava\\OneDrive - Shell\\Apps\\ml-comm\\smile\\facerecog\\FaceRecognition-master\\ids\\" + str(
count) + "\\"
img_name = path_img + str(count) + '.png'
cv2.imwrite(img_name, frame)
# start = time.time()
entry_time = datetime.datetime.fromtimestamp(
start).strftime('%c')
df_tmp = pd.DataFrame(
[[str(count), entry_time, "Still_Inside"]],
columns=df.columns)
count = count + 1
count_msg = "Evacuated staff: " + str(count)
remaining_msg = "Remaining staff: " + str(
total_people - count)
df = df.append(df_tmp)
df.to_csv(f, header=False, index=False)
if show_id:
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(frame, matching_id, (bb[0], bb[3]),
font, 1, (255, 255, 255), 1,
cv2.LINE_AA)
if show_bb:
cv2.rectangle(frame, (bb[0], bb[1]),
(bb[2], bb[3]), (255, 0, 0), 2)
if show_landmarks:
for j in range(5):
size = 1
top_left = (int(landmarks[i, j]) - size,
int(landmarks[i, j + 5]) - size)
bottom_right = (int(landmarks[i, j]) + size,
int(landmarks[i, j + 5]) +
size)
cv2.rectangle(frame, top_left, bottom_right,
(255, 0, 255), 2)
else:
print('Couldn\'t find a face')
if len(ids) < 5:
show_missing(frames, ids, 0, int_frame_height)
cv2.putText(frames, remaining_msg, (0, int_frame_height - 15),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 1,
cv2.LINE_AA)
cv2.putText(frames, count_msg, (0, 145),
cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 255, 0), 2,
cv2.LINE_AA)
end = time.time()
seconds = end - start
fps = round(1 / seconds, 2)
if show_fps:
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(frames, str(fps), (0, int(frame_height) - 5),
font, 1, (255, 255, 255), 1, cv2.LINE_AA)
cv2.imshow('frame', frames)
key = cv2.waitKey(1)
if key == ord('q'):
f.close()
break
elif key == ord('l'):
show_landmarks = not show_landmarks
elif key == ord('b'):
show_bb = not show_bb
elif key == ord('i'):
show_id = not show_id
elif key == ord('f'):
show_fps = not show_fps
cap.release()
cv2.destroyAllWindows()
