github/deepface
1
1
Fork 0
mirror of https://github.com/serengil/deepface.git synced 2025-06-09 21:07:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

Pass custom path argument while loading model

Browse Source 
Add `model_path` parameter in `loadModel()` function. This adds more
flexiblity while loading the models.

Also, refactor code using standard `os.path.join` which will make sure
to join the paths correctly.
This commit is contained in:
NISH1001 2020-05-20 15:55:08 +05:45
parent 892cbae56a
commit 2c0a507994
7 changed files with 3448 additions and 1447 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

646
deepface/DeepFace.py
View File

@ -1,5 +1,6 @@
from keras.preprocessing import image from keras.preprocessing import image
import warnings import warnings
warnings.filterwarnings("ignore") warnings.filterwarnings("ignore")
import time import time
import os import os
@ -12,330 +13,397 @@ from keras import backend as K
import keras import keras
import tensorflow as tf import tensorflow as tf
#from basemodels import VGGFace, OpenFace, Facenet, FbDeepFace # from basemodels import VGGFace, OpenFace, Facenet, FbDeepFace
#from extendedmodels import Age, Gender, Race, Emotion # from extendedmodels import Age, Gender, Race, Emotion
#from commons import functions, realtime, distance as dst # from commons import functions, realtime, distance as dst
from deepface.basemodels import VGGFace, OpenFace, Facenet, FbDeepFace from deepface.basemodels import VGGFace, OpenFace, Facenet, FbDeepFace
from deepface.extendedmodels import Age, Gender, Race, Emotion from deepface.extendedmodels import Age, Gender, Race, Emotion
from deepface.commons import functions, realtime, distance as dst from deepface.commons import functions, realtime, distance as dst
def verify(img1_path, img2_path=''
, model_name ='VGG-Face', distance_metric = 'cosine', model = None, enforce_detection = True):
tic = time.time() def verify(
img1_path,
img2_path="",
model_name="VGG-Face",
distance_metric="cosine",
model=None,
enforce_detection=True,
):
tic = time.time()
if type(img1_path) == list:
bulkProcess = True
img_list = img1_path.copy()
else:
bulkProcess = False
img_list = [[img1_path, img2_path]]
# ------------------------------
if model == None:
if model_name == "VGG-Face":
print("Using VGG-Face model backend and", distance_metric, "distance.")
model = VGGFace.loadModel()
elif model_name == "OpenFace":
print("Using OpenFace model backend", distance_metric, "distance.")
model = OpenFace.loadModel()
elif model_name == "Facenet":
print("Using Facenet model backend", distance_metric, "distance.")
model = Facenet.loadModel()
elif model_name == "DeepFace":
print("Using FB DeepFace model backend", distance_metric, "distance.")
model = FbDeepFace.loadModel()
if type(img1_path) == list: else:
bulkProcess = True raise ValueError("Invalid model_name passed - ", model_name)
img_list = img1_path.copy() else: # model != None
else: print("Already built model is passed")
bulkProcess = False
img_list = [[img1_path, img2_path]]
#------------------------------ # ------------------------------
# face recognition models have different size of inputs
input_shape = model.layers[0].input_shape[1:3]
if model == None: # ------------------------------
if model_name == 'VGG-Face':
print("Using VGG-Face model backend and", distance_metric,"distance.")
model = VGGFace.loadModel()
elif model_name == 'OpenFace': # tuned thresholds for model and metric pair
print("Using OpenFace model backend", distance_metric,"distance.") threshold = functions.findThreshold(model_name, distance_metric)
model = OpenFace.loadModel()
# ------------------------------
pbar = tqdm(range(0, len(img_list)), desc="Verification")
elif model_name == 'Facenet': resp_objects = []
print("Using Facenet model backend", distance_metric,"distance.")
model = Facenet.loadModel()
elif model_name == 'DeepFace': # for instance in img_list:
print("Using FB DeepFace model backend", distance_metric,"distance.") for index in pbar:
model = FbDeepFace.loadModel()
else: instance = img_list[index]
raise ValueError("Invalid model_name passed - ", model_name)
else: #model != None
print("Already built model is passed")
#------------------------------ if type(instance) == list and len(instance) >= 2:
#face recognition models have different size of inputs img1_path = instance[0]
input_shape = model.layers[0].input_shape[1:3] img2_path = instance[1]
#------------------------------ # ----------------------
# crop and align faces
#tuned thresholds for model and metric pair img1 = functions.detectFace(
threshold = functions.findThreshold(model_name, distance_metric) img1_path, input_shape, enforce_detection=enforce_detection
)
img2 = functions.detectFace(
img2_path, input_shape, enforce_detection=enforce_detection
)
#------------------------------ # ----------------------
pbar = tqdm(range(0,len(img_list)), desc='Verification') # find embeddings
resp_objects = [] img1_representation = model.predict(img1)[0, :]
img2_representation = model.predict(img2)[0, :]
#for instance in img_list: # ----------------------
for index in pbar: # find distances between embeddings
instance = img_list[index] if distance_metric == "cosine":
distance = dst.findCosineDistance(
img1_representation, img2_representation
)
elif distance_metric == "euclidean":
distance = dst.findEuclideanDistance(
img1_representation, img2_representation
)
elif distance_metric == "euclidean_l2":
distance = dst.findEuclideanDistance(
dst.l2_normalize(img1_representation),
dst.l2_normalize(img2_representation),
)
else:
raise ValueError("Invalid distance_metric passed - ", distance_metric)
# ----------------------
# decision
if distance <= threshold:
identified = "true"
else:
identified = "false"
if type(instance) == list and len(instance) >= 2: # ----------------------
img1_path = instance[0] # response object
img2_path = instance[1]
#---------------------- resp_obj = "{"
#crop and align faces resp_obj += '"verified": ' + identified
resp_obj += ', "distance": ' + str(distance)
resp_obj += ', "max_threshold_to_verify": ' + str(threshold)
resp_obj += ', "model": "' + model_name + '"'
resp_obj += ', "similarity_metric": "' + distance_metric + '"'
resp_obj += "}"
img1 = functions.detectFace(img1_path, input_shape, enforce_detection = enforce_detection) resp_obj = json.loads(resp_obj) # string to json
img2 = functions.detectFace(img2_path, input_shape, enforce_detection = enforce_detection)
#---------------------- if bulkProcess == True:
#find embeddings resp_objects.append(resp_obj)
else:
# K.clear_session()
return resp_obj
# ----------------------
img1_representation = model.predict(img1)[0,:] else:
img2_representation = model.predict(img2)[0,:] raise ValueError("Invalid arguments passed to verify function: ", instance)
#---------------------- # -------------------------
#find distances between embeddings
if distance_metric == 'cosine': toc = time.time()
distance = dst.findCosineDistance(img1_representation, img2_representation)
elif distance_metric == 'euclidean': # print("identification lasts ",toc-tic," seconds")
distance = dst.findEuclideanDistance(img1_representation, img2_representation)
elif distance_metric == 'euclidean_l2': if bulkProcess == True:
distance = dst.findEuclideanDistance(dst.l2_normalize(img1_representation), dst.l2_normalize(img2_representation)) resp_obj = "{"
else:
raise ValueError("Invalid distance_metric passed - ", distance_metric) for i in range(0, len(resp_objects)):
resp_item = json.dumps(resp_objects[i])
if i > 0:
resp_obj += ", "
#---------------------- resp_obj += '"pair_' + str(i + 1) + '": ' + resp_item
#decision resp_obj += "}"
resp_obj = json.loads(resp_obj)
if distance <= threshold: return resp_obj
identified = "true" # return resp_objects
else:
identified = "false"
def analyze(img_path, actions=[], models={}, enforce_detection=True):
#----------------------
#response object if type(img_path) == list:
img_paths = img_path.copy()
resp_obj = "{" bulkProcess = True
resp_obj += "\"verified\": "+identified else:
resp_obj += ", \"distance\": "+str(distance) img_paths = [img_path]
resp_obj += ", \"max_threshold_to_verify\": "+str(threshold) bulkProcess = False
resp_obj += ", \"model\": \""+model_name+"\""
resp_obj += ", \"similarity_metric\": \""+distance_metric+"\"" # ---------------------------------
resp_obj += "}"
# if a specific target is not passed, then find them all
resp_obj = json.loads(resp_obj) #string to json if len(actions) == 0:
actions = ["emotion", "age", "gender", "race"]
if bulkProcess == True:
resp_objects.append(resp_obj) print("Actions to do: ", actions)
else:
#K.clear_session() # ---------------------------------
return resp_obj
#---------------------- if "emotion" in actions:
if "emotion" in models:
else: print("already built emotion model is passed")
raise ValueError("Invalid arguments passed to verify function: ", instance) emotion_model = models["emotion"]
else:
#------------------------- emotion_model = Emotion.loadModel()
toc = time.time() if "age" in actions:
if "age" in models:
#print("identification lasts ",toc-tic," seconds") print("already built age model is passed")
age_model = models["age"]
if bulkProcess == True: else:
resp_obj = "{" age_model = Age.loadModel()
for i in range(0, len(resp_objects)): if "gender" in actions:
resp_item = json.dumps(resp_objects[i]) if "gender" in models:
print("already built gender model is passed")
if i > 0: gender_model = models["gender"]
resp_obj += ", " else:
gender_model = Gender.loadModel()
resp_obj += "\"pair_"+str(i+1)+"\": "+resp_item
resp_obj += "}" if "race" in actions:
resp_obj = json.loads(resp_obj) if "race" in models:
return resp_obj print("already built race model is passed")
#return resp_objects race_model = models["race"]
else:
race_model = Race.loadModel()
def analyze(img_path, actions = [], models = {}, enforce_detection = True): # ---------------------------------
if type(img_path) == list: resp_objects = []
img_paths = img_path.copy()
bulkProcess = True global_pbar = tqdm(range(0, len(img_paths)), desc="Analyzing")
else:
img_paths = [img_path] # for img_path in img_paths:
bulkProcess = False for j in global_pbar:
img_path = img_paths[j]
#---------------------------------
resp_obj = "{"
#if a specific target is not passed, then find them all
if len(actions) == 0: # TO-DO: do this in parallel
actions= ['emotion', 'age', 'gender', 'race']
pbar = tqdm(range(0, len(actions)), desc="Finding actions")
print("Actions to do: ", actions)
action_idx = 0
#--------------------------------- img_224 = None # Set to prevent re-detection
# for action in actions:
if 'emotion' in actions: for index in pbar:
if 'emotion' in models: action = actions[index]
print("already built emotion model is passed") pbar.set_description("Action: %s" % (action))
emotion_model = models['emotion']
else: if action_idx > 0:
emotion_model = Emotion.loadModel() resp_obj += ", "
if 'age' in actions: if action == "emotion":
if 'age' in models: emotion_labels = [
print("already built age model is passed") "angry",
age_model = models['age'] "disgust",
else: "fear",
age_model = Age.loadModel() "happy",
"sad",
if 'gender' in actions: "surprise",
if 'gender' in models: "neutral",
print("already built gender model is passed") ]
gender_model = models['gender'] img = functions.detectFace(
else: img_path,
gender_model = Gender.loadModel() target_size=(48, 48),
grayscale=True,
if 'race' in actions: enforce_detection=enforce_detection,
if 'race' in models: )
print("already built race model is passed")
race_model = models['race'] emotion_predictions = emotion_model.predict(img)[0, :]
else:
race_model = Race.loadModel() sum_of_predictions = emotion_predictions.sum()
#---------------------------------
emotion_obj = '"emotion": {'
resp_objects = [] for i in range(0, len(emotion_labels)):
emotion_label = emotion_labels[i]
global_pbar = tqdm(range(0,len(img_paths)), desc='Analyzing') emotion_prediction = (
100 * emotion_predictions[i] / sum_of_predictions
#for img_path in img_paths: )
for j in global_pbar:
img_path = img_paths[j] if i > 0:
emotion_obj += ", "
resp_obj = "{"
emotion_obj += '"%s": %s' % (emotion_label, emotion_prediction)
#TO-DO: do this in parallel
emotion_obj += "}"
pbar = tqdm(range(0,len(actions)), desc='Finding actions')
emotion_obj += ', "dominant_emotion": "%s"' % (
action_idx = 0 emotion_labels[np.argmax(emotion_predictions)]
img_224 = None # Set to prevent re-detection )
#for action in actions:
for index in pbar: resp_obj += emotion_obj
action = actions[index]
pbar.set_description("Action: %s" % (action)) elif action == "age":
if img_224 is None:
if action_idx > 0: img_224 = functions.detectFace(
resp_obj += ", " img_path,
target_size=(224, 224),
if action == 'emotion': grayscale=False,
emotion_labels = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] enforce_detection=enforce_detection,
img = functions.detectFace(img_path, target_size = (48, 48), grayscale = True, enforce_detection = enforce_detection) ) # just emotion model expects grayscale images
# print("age prediction")
emotion_predictions = emotion_model.predict(img)[0,:] age_predictions = age_model.predict(img_224)[0, :]
apparent_age = Age.findApparentAge(age_predictions)
sum_of_predictions = emotion_predictions.sum()
resp_obj += '"age": %s' % (apparent_age)
emotion_obj = "\"emotion\": {"
for i in range(0, len(emotion_labels)): elif action == "gender":
emotion_label = emotion_labels[i] if img_224 is None:
emotion_prediction = 100 * emotion_predictions[i] / sum_of_predictions img_224 = functions.detectFace(
img_path,
if i > 0: emotion_obj += ", " target_size=(224, 224),
grayscale=False,
emotion_obj += "\"%s\": %s" % (emotion_label, emotion_prediction) enforce_detection=enforce_detection,
) # just emotion model expects grayscale images
emotion_obj += "}" # print("gender prediction")
emotion_obj += ", \"dominant_emotion\": \"%s\"" % (emotion_labels[np.argmax(emotion_predictions)]) gender_prediction = gender_model.predict(img_224)[0, :]
resp_obj += emotion_obj if np.argmax(gender_prediction) == 0:
gender = "Woman"
elif action == 'age': elif np.argmax(gender_prediction) == 1:
if img_224 is None: gender = "Man"
img_224 = functions.detectFace(img_path, target_size = (224, 224), grayscale = False, enforce_detection = enforce_detection) #just emotion model expects grayscale images
#print("age prediction") resp_obj += '"gender": "%s"' % (gender)
age_predictions = age_model.predict(img_224)[0,:]
apparent_age = Age.findApparentAge(age_predictions) elif action == "race":
if img_224 is None:
resp_obj += "\"age\": %s" % (apparent_age) img_224 = functions.detectFace(
img_path,
elif action == 'gender': target_size=(224, 224),
if img_224 is None: grayscale=False,
img_224 = functions.detectFace(img_path, target_size = (224, 224), grayscale = False, enforce_detection = enforce_detection) #just emotion model expects grayscale images enforce_detection=enforce_detection,
#print("gender prediction") ) # just emotion model expects grayscale images
race_predictions = race_model.predict(img_224)[0, :]
gender_prediction = gender_model.predict(img_224)[0,:] race_labels = [
"asian",
if np.argmax(gender_prediction) == 0: "indian",
gender = "Woman" "black",
elif np.argmax(gender_prediction) == 1: "white",
gender = "Man" "middle eastern",
"latino hispanic",
resp_obj += "\"gender\": \"%s\"" % (gender) ]
elif action == 'race': sum_of_predictions = race_predictions.sum()
if img_224 is None:
img_224 = functions.detectFace(img_path, target_size = (224, 224), grayscale = False, enforce_detection = enforce_detection) #just emotion model expects grayscale images race_obj = '"race": {'
race_predictions = race_model.predict(img_224)[0,:] for i in range(0, len(race_labels)):
race_labels = ['asian', 'indian', 'black', 'white', 'middle eastern', 'latino hispanic'] race_label = race_labels[i]
race_prediction = 100 * race_predictions[i] / sum_of_predictions
sum_of_predictions = race_predictions.sum()
if i > 0:
race_obj = "\"race\": {" race_obj += ", "
for i in range(0, len(race_labels)):
race_label = race_labels[i] race_obj += '"%s": %s' % (race_label, race_prediction)
race_prediction = 100 * race_predictions[i] / sum_of_predictions
race_obj += "}"
if i > 0: race_obj += ", " race_obj += ', "dominant_race": "%s"' % (
race_labels[np.argmax(race_predictions)]
race_obj += "\"%s\": %s" % (race_label, race_prediction) )
race_obj += "}" resp_obj += race_obj
race_obj += ", \"dominant_race\": \"%s\"" % (race_labels[np.argmax(race_predictions)])
action_idx = action_idx + 1
resp_obj += race_obj
resp_obj += "}"
action_idx = action_idx + 1
resp_obj = json.loads(resp_obj)
resp_obj += "}"
if bulkProcess == True:
resp_obj = json.loads(resp_obj) resp_objects.append(resp_obj)
else:
if bulkProcess == True: return resp_obj
resp_objects.append(resp_obj)
else: if bulkProcess == True:
return resp_obj resp_obj = "{"
if bulkProcess == True: for i in range(0, len(resp_objects)):
resp_obj = "{" resp_item = json.dumps(resp_objects[i])
for i in range(0, len(resp_objects)): if i > 0:
resp_item = json.dumps(resp_objects[i]) resp_obj += ", "
if i > 0: resp_obj += '"instance_' + str(i + 1) + '": ' + resp_item
resp_obj += ", " resp_obj += "}"
resp_obj = json.loads(resp_obj)
resp_obj += "\"instance_"+str(i+1)+"\": "+resp_item return resp_obj
resp_obj += "}" # return resp_objects
resp_obj = json.loads(resp_obj)
return resp_obj
#return resp_objects
def detectFace(img_path): def detectFace(img_path):
img = functions.detectFace(img_path)[0] #detectFace returns (1, 224, 224, 3) img = functions.detectFace(img_path)[0] # detectFace returns (1, 224, 224, 3)
return img[:, :, ::-1] #bgr to rgb return img[:, :, ::-1] # bgr to rgb
def stream(db_path = '', model_name ='VGG-Face', distance_metric = 'cosine', enable_face_analysis = True): def stream(
realtime.analysis(db_path, model_name, distance_metric, enable_face_analysis) db_path="",
model_name="VGG-Face",
distance_metric="cosine",
enable_face_analysis=True,
):
realtime.analysis(db_path, model_name, distance_metric, enable_face_analysis)
def allocateMemory(): def allocateMemory():
print("Analyzing your system...") print("Analyzing your system...")
functions.allocateMemory() functions.allocateMemory()
functions.initializeFolder() functions.initializeFolder()
#--------------------------- # ---------------------------

2678
deepface/basemodels/Facenet.py
View File

File diff suppressed because it is too large Load Diff

85
deepface/basemodels/FbDeepFace.py
View File

@ -3,44 +3,71 @@ from pathlib import Path
import gdown import gdown
import keras import keras
from keras.models import Model, Sequential from keras.models import Model, Sequential
from keras.layers import Convolution2D, LocallyConnected2D, MaxPooling2D, Flatten, Dense, Dropout from keras.layers import (
Convolution2D,
LocallyConnected2D,
MaxPooling2D,
Flatten,
Dense,
Dropout,
)
import zipfile import zipfile
#------------------------------------- # -------------------------------------
def loadModel():
base_model = Sequential()
base_model.add(Convolution2D(32, (11, 11), activation='relu', name='C1', input_shape=(152, 152, 3)))
base_model.add(MaxPooling2D(pool_size=3, strides=2, padding='same', name='M2'))
base_model.add(Convolution2D(16, (9, 9), activation='relu', name='C3'))
base_model.add(LocallyConnected2D(16, (9, 9), activation='relu', name='L4'))
base_model.add(LocallyConnected2D(16, (7, 7), strides=2, activation='relu', name='L5') )
base_model.add(LocallyConnected2D(16, (5, 5), activation='relu', name='L6'))
base_model.add(Flatten(name='F0'))
base_model.add(Dense(4096, activation='relu', name='F7'))
base_model.add(Dropout(rate=0.5, name='D0'))
base_model.add(Dense(8631, activation='softmax', name='F8'))
#--------------------------------- def get_base_model():
base_model = Sequential()
base_model.add(
Convolution2D(
32, (11, 11), activation="relu", name="C1", input_shape=(152, 152, 3)
)
)
base_model.add(MaxPooling2D(pool_size=3, strides=2, padding="same", name="M2"))
base_model.add(Convolution2D(16, (9, 9), activation="relu", name="C3"))
base_model.add(LocallyConnected2D(16, (9, 9), activation="relu", name="L4"))
base_model.add(
LocallyConnected2D(16, (7, 7), strides=2, activation="relu", name="L5")
)
base_model.add(LocallyConnected2D(16, (5, 5), activation="relu", name="L6"))
base_model.add(Flatten(name="F0"))
base_model.add(Dense(4096, activation="relu", name="F7"))
base_model.add(Dropout(rate=0.5, name="D0"))
base_model.add(Dense(8631, activation="softmax", name="F8"))
return base_model
home = str(Path.home())
if os.path.isfile(home+'/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5') != True: def loadModel(model_path=""):
print("VGGFace2_DeepFace_weights_val-0.9034.h5 will be downloaded...") # ---------------------------------
if model_path:
assert Path(model_path).exists()
assert model_path.endswith(".h5")
else:
home = Path.home().as_posix()
model_path = os.path.join(
home, ".deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5"
)
if not os.path.isfile(model_path):
print("VGGFace2_DeepFace_weights_val-0.9034.h5 will be downloaded...")
url = 'https://github.com/swghosh/DeepFace/releases/download/weights-vggface2-2d-aligned/VGGFace2_DeepFace_weights_val-0.9034.h5.zip' url = "https://github.com/swghosh/DeepFace/releases/download/weights-vggface2-2d-aligned/VGGFace2_DeepFace_weights_val-0.9034.h5.zip"
output = home+'/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5.zip' zip_path = os.path.join(
home, ".deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5.zip"
)
gdown.download(url, zip_path, quiet=False)
gdown.download(url, output, quiet=False) # unzip VGGFace2_DeepFace_weights_val-0.9034.h5.zip
with zipfile.ZipFile(zip_path, "r") as zip_ref:
zip_ref.extractall(os.path.join(home, "/.deepface/weights/"))
#unzip VGGFace2_DeepFace_weights_val-0.9034.h5.zip print(f"Loading model from {model_path}")
with zipfile.ZipFile(output, 'r') as zip_ref: base_model = get_base_model()
zip_ref.extractall(home+'/.deepface/weights/') base_model.load_weights(model_path)
base_model.load_weights(home+'/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5') # drop F8 and D0. F7 is the representation layer.
deepface_model = Model(
inputs=base_model.layers[0].input, outputs=base_model.layers[-3].output
)
#drop F8 and D0. F7 is the representation layer. return deepface_model
deepface_model = Model(inputs=base_model.layers[0].input, outputs=base_model.layers[-3].output)
return deepface_model

535
deepface/basemodels/OpenFace.py
View File

@ -13,238 +13,397 @@ from keras.layers.normalization import BatchNormalization
from keras.models import load_model from keras.models import load_model
from keras import backend as K from keras import backend as K
#--------------------------------------- # ---------------------------------------
def loadModel():
myInput = Input(shape=(96, 96, 3))
x = ZeroPadding2D(padding=(3, 3), input_shape=(96, 96, 3))(myInput) def get_base_model():
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1')(x) myInput = Input(shape=(96, 96, 3))
x = BatchNormalization(axis=3, epsilon=0.00001, name='bn1')(x)
x = Activation('relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = MaxPooling2D(pool_size=3, strides=2)(x)
x = Lambda(lambda x: tf.nn.lrn(x, alpha=1e-4, beta=0.75), name='lrn_1')(x)
x = Conv2D(64, (1, 1), name='conv2')(x)
x = BatchNormalization(axis=3, epsilon=0.00001, name='bn2')(x)
x = Activation('relu')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(192, (3, 3), name='conv3')(x)
x = BatchNormalization(axis=3, epsilon=0.00001, name='bn3')(x)
x = Activation('relu')(x)
x = Lambda(lambda x: tf.nn.lrn(x, alpha=1e-4, beta=0.75), name='lrn_2')(x) #x is equal added
x = ZeroPadding2D(padding=(1, 1))(x)
x = MaxPooling2D(pool_size=3, strides=2)(x)
# Inception3a x = ZeroPadding2D(padding=(3, 3), input_shape=(96, 96, 3))(myInput)
inception_3a_3x3 = Conv2D(96, (1, 1), name='inception_3a_3x3_conv1')(x) x = Conv2D(64, (7, 7), strides=(2, 2), name="conv1")(x)
inception_3a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_3x3_bn1')(inception_3a_3x3) x = BatchNormalization(axis=3, epsilon=0.00001, name="bn1")(x)
inception_3a_3x3 = Activation('relu')(inception_3a_3x3) x = Activation("relu")(x)
inception_3a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3a_3x3) x = ZeroPadding2D(padding=(1, 1))(x)
inception_3a_3x3 = Conv2D(128, (3, 3), name='inception_3a_3x3_conv2')(inception_3a_3x3) x = MaxPooling2D(pool_size=3, strides=2)(x)
inception_3a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_3x3_bn2')(inception_3a_3x3) x = Lambda(lambda x: tf.nn.lrn(x, alpha=1e-4, beta=0.75), name="lrn_1")(x)
inception_3a_3x3 = Activation('relu')(inception_3a_3x3) x = Conv2D(64, (1, 1), name="conv2")(x)
x = BatchNormalization(axis=3, epsilon=0.00001, name="bn2")(x)
x = Activation("relu")(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = Conv2D(192, (3, 3), name="conv3")(x)
x = BatchNormalization(axis=3, epsilon=0.00001, name="bn3")(x)
x = Activation("relu")(x)
x = Lambda(lambda x: tf.nn.lrn(x, alpha=1e-4, beta=0.75), name="lrn_2")(
x
) # x is equal added
x = ZeroPadding2D(padding=(1, 1))(x)
x = MaxPooling2D(pool_size=3, strides=2)(x)
inception_3a_5x5 = Conv2D(16, (1, 1), name='inception_3a_5x5_conv1')(x) # Inception3a
inception_3a_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_5x5_bn1')(inception_3a_5x5) inception_3a_3x3 = Conv2D(96, (1, 1), name="inception_3a_3x3_conv1")(x)
inception_3a_5x5 = Activation('relu')(inception_3a_5x5) inception_3a_3x3 = BatchNormalization(
inception_3a_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3a_5x5) axis=3, epsilon=0.00001, name="inception_3a_3x3_bn1"
inception_3a_5x5 = Conv2D(32, (5, 5), name='inception_3a_5x5_conv2')(inception_3a_5x5) )(inception_3a_3x3)
inception_3a_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_5x5_bn2')(inception_3a_5x5) inception_3a_3x3 = Activation("relu")(inception_3a_3x3)
inception_3a_5x5 = Activation('relu')(inception_3a_5x5) inception_3a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3a_3x3)
inception_3a_3x3 = Conv2D(128, (3, 3), name="inception_3a_3x3_conv2")(
inception_3a_3x3
)
inception_3a_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_3x3_bn2"
)(inception_3a_3x3)
inception_3a_3x3 = Activation("relu")(inception_3a_3x3)
inception_3a_pool = MaxPooling2D(pool_size=3, strides=2)(x) inception_3a_5x5 = Conv2D(16, (1, 1), name="inception_3a_5x5_conv1")(x)
inception_3a_pool = Conv2D(32, (1, 1), name='inception_3a_pool_conv')(inception_3a_pool) inception_3a_5x5 = BatchNormalization(
inception_3a_pool = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_pool_bn')(inception_3a_pool) axis=3, epsilon=0.00001, name="inception_3a_5x5_bn1"
inception_3a_pool = Activation('relu')(inception_3a_pool) )(inception_3a_5x5)
inception_3a_pool = ZeroPadding2D(padding=((3, 4), (3, 4)))(inception_3a_pool) inception_3a_5x5 = Activation("relu")(inception_3a_5x5)
inception_3a_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3a_5x5)
inception_3a_5x5 = Conv2D(32, (5, 5), name="inception_3a_5x5_conv2")(
inception_3a_5x5
)
inception_3a_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_5x5_bn2"
)(inception_3a_5x5)
inception_3a_5x5 = Activation("relu")(inception_3a_5x5)
inception_3a_1x1 = Conv2D(64, (1, 1), name='inception_3a_1x1_conv')(x) inception_3a_pool = MaxPooling2D(pool_size=3, strides=2)(x)
inception_3a_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_1x1_bn')(inception_3a_1x1) inception_3a_pool = Conv2D(32, (1, 1), name="inception_3a_pool_conv")(
inception_3a_1x1 = Activation('relu')(inception_3a_1x1) inception_3a_pool
)
inception_3a_pool = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_pool_bn"
)(inception_3a_pool)
inception_3a_pool = Activation("relu")(inception_3a_pool)
inception_3a_pool = ZeroPadding2D(padding=((3, 4), (3, 4)))(inception_3a_pool)
inception_3a = concatenate([inception_3a_3x3, inception_3a_5x5, inception_3a_pool, inception_3a_1x1], axis=3) inception_3a_1x1 = Conv2D(64, (1, 1), name="inception_3a_1x1_conv")(x)
inception_3a_1x1 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_1x1_bn"
)(inception_3a_1x1)
inception_3a_1x1 = Activation("relu")(inception_3a_1x1)
# Inception3b inception_3a = concatenate(
inception_3b_3x3 = Conv2D(96, (1, 1), name='inception_3b_3x3_conv1')(inception_3a) [inception_3a_3x3, inception_3a_5x5, inception_3a_pool, inception_3a_1x1],
inception_3b_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_3x3_bn1')(inception_3b_3x3) axis=3,
inception_3b_3x3 = Activation('relu')(inception_3b_3x3) )
inception_3b_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3b_3x3)
inception_3b_3x3 = Conv2D(128, (3, 3), name='inception_3b_3x3_conv2')(inception_3b_3x3)
inception_3b_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_3x3_bn2')(inception_3b_3x3)
inception_3b_3x3 = Activation('relu')(inception_3b_3x3)
inception_3b_5x5 = Conv2D(32, (1, 1), name='inception_3b_5x5_conv1')(inception_3a) # Inception3b
inception_3b_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_5x5_bn1')(inception_3b_5x5) inception_3b_3x3 = Conv2D(96, (1, 1), name="inception_3b_3x3_conv1")(inception_3a)
inception_3b_5x5 = Activation('relu')(inception_3b_5x5) inception_3b_3x3 = BatchNormalization(
inception_3b_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3b_5x5) axis=3, epsilon=0.00001, name="inception_3b_3x3_bn1"
inception_3b_5x5 = Conv2D(64, (5, 5), name='inception_3b_5x5_conv2')(inception_3b_5x5) )(inception_3b_3x3)
inception_3b_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_5x5_bn2')(inception_3b_5x5) inception_3b_3x3 = Activation("relu")(inception_3b_3x3)
inception_3b_5x5 = Activation('relu')(inception_3b_5x5) inception_3b_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3b_3x3)
inception_3b_3x3 = Conv2D(128, (3, 3), name="inception_3b_3x3_conv2")(
inception_3b_3x3
)
inception_3b_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_3x3_bn2"
)(inception_3b_3x3)
inception_3b_3x3 = Activation("relu")(inception_3b_3x3)
inception_3b_pool = Lambda(lambda x: x**2, name='power2_3b')(inception_3a) inception_3b_5x5 = Conv2D(32, (1, 1), name="inception_3b_5x5_conv1")(inception_3a)
inception_3b_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(inception_3b_pool) inception_3b_5x5 = BatchNormalization(
inception_3b_pool = Lambda(lambda x: x*9, name='mult9_3b')(inception_3b_pool) axis=3, epsilon=0.00001, name="inception_3b_5x5_bn1"
inception_3b_pool = Lambda(lambda x: K.sqrt(x), name='sqrt_3b')(inception_3b_pool) )(inception_3b_5x5)
inception_3b_pool = Conv2D(64, (1, 1), name='inception_3b_pool_conv')(inception_3b_pool) inception_3b_5x5 = Activation("relu")(inception_3b_5x5)
inception_3b_pool = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_pool_bn')(inception_3b_pool) inception_3b_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3b_5x5)
inception_3b_pool = Activation('relu')(inception_3b_pool) inception_3b_5x5 = Conv2D(64, (5, 5), name="inception_3b_5x5_conv2")(
inception_3b_pool = ZeroPadding2D(padding=(4, 4))(inception_3b_pool) inception_3b_5x5
)
inception_3b_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_5x5_bn2"
)(inception_3b_5x5)
inception_3b_5x5 = Activation("relu")(inception_3b_5x5)
inception_3b_1x1 = Conv2D(64, (1, 1), name='inception_3b_1x1_conv')(inception_3a) inception_3b_pool = Lambda(lambda x: x ** 2, name="power2_3b")(inception_3a)
inception_3b_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_1x1_bn')(inception_3b_1x1) inception_3b_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
inception_3b_1x1 = Activation('relu')(inception_3b_1x1) inception_3b_pool
)
inception_3b_pool = Lambda(lambda x: x * 9, name="mult9_3b")(inception_3b_pool)
inception_3b_pool = Lambda(lambda x: K.sqrt(x), name="sqrt_3b")(inception_3b_pool)
inception_3b_pool = Conv2D(64, (1, 1), name="inception_3b_pool_conv")(
inception_3b_pool
)
inception_3b_pool = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_pool_bn"
)(inception_3b_pool)
inception_3b_pool = Activation("relu")(inception_3b_pool)
inception_3b_pool = ZeroPadding2D(padding=(4, 4))(inception_3b_pool)
inception_3b = concatenate([inception_3b_3x3, inception_3b_5x5, inception_3b_pool, inception_3b_1x1], axis=3) inception_3b_1x1 = Conv2D(64, (1, 1), name="inception_3b_1x1_conv")(inception_3a)
inception_3b_1x1 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_1x1_bn"
)(inception_3b_1x1)
inception_3b_1x1 = Activation("relu")(inception_3b_1x1)
# Inception3c inception_3b = concatenate(
inception_3c_3x3 = Conv2D(128, (1, 1), strides=(1, 1), name='inception_3c_3x3_conv1')(inception_3b) [inception_3b_3x3, inception_3b_5x5, inception_3b_pool, inception_3b_1x1],
inception_3c_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3c_3x3_bn1')(inception_3c_3x3) axis=3,
inception_3c_3x3 = Activation('relu')(inception_3c_3x3) )
inception_3c_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3c_3x3)
inception_3c_3x3 = Conv2D(256, (3, 3), strides=(2, 2), name='inception_3c_3x3_conv'+'2')(inception_3c_3x3)
inception_3c_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3c_3x3_bn'+'2')(inception_3c_3x3)
inception_3c_3x3 = Activation('relu')(inception_3c_3x3)
inception_3c_5x5 = Conv2D(32, (1, 1), strides=(1, 1), name='inception_3c_5x5_conv1')(inception_3b) # Inception3c
inception_3c_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3c_5x5_bn1')(inception_3c_5x5) inception_3c_3x3 = Conv2D(
inception_3c_5x5 = Activation('relu')(inception_3c_5x5) 128, (1, 1), strides=(1, 1), name="inception_3c_3x3_conv1"
inception_3c_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3c_5x5) )(inception_3b)
inception_3c_5x5 = Conv2D(64, (5, 5), strides=(2, 2), name='inception_3c_5x5_conv'+'2')(inception_3c_5x5) inception_3c_3x3 = BatchNormalization(
inception_3c_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3c_5x5_bn'+'2')(inception_3c_5x5) axis=3, epsilon=0.00001, name="inception_3c_3x3_bn1"
inception_3c_5x5 = Activation('relu')(inception_3c_5x5) )(inception_3c_3x3)
inception_3c_3x3 = Activation("relu")(inception_3c_3x3)
inception_3c_3x3 = ZeroPadding2D(padding=(1, 1))(inception_3c_3x3)
inception_3c_3x3 = Conv2D(
256, (3, 3), strides=(2, 2), name="inception_3c_3x3_conv" + "2"
)(inception_3c_3x3)
inception_3c_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3c_3x3_bn" + "2"
)(inception_3c_3x3)
inception_3c_3x3 = Activation("relu")(inception_3c_3x3)
inception_3c_pool = MaxPooling2D(pool_size=3, strides=2)(inception_3b) inception_3c_5x5 = Conv2D(
inception_3c_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_3c_pool) 32, (1, 1), strides=(1, 1), name="inception_3c_5x5_conv1"
)(inception_3b)
inception_3c_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3c_5x5_bn1"
)(inception_3c_5x5)
inception_3c_5x5 = Activation("relu")(inception_3c_5x5)
inception_3c_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3c_5x5)
inception_3c_5x5 = Conv2D(
64, (5, 5), strides=(2, 2), name="inception_3c_5x5_conv" + "2"
)(inception_3c_5x5)
inception_3c_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3c_5x5_bn" + "2"
)(inception_3c_5x5)
inception_3c_5x5 = Activation("relu")(inception_3c_5x5)
inception_3c = concatenate([inception_3c_3x3, inception_3c_5x5, inception_3c_pool], axis=3) inception_3c_pool = MaxPooling2D(pool_size=3, strides=2)(inception_3b)
inception_3c_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_3c_pool)
#inception 4a inception_3c = concatenate(
inception_4a_3x3 = Conv2D(96, (1, 1), strides=(1, 1), name='inception_4a_3x3_conv'+'1')(inception_3c) [inception_3c_3x3, inception_3c_5x5, inception_3c_pool], axis=3
inception_4a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_3x3_bn'+'1')(inception_4a_3x3) )
inception_4a_3x3 = Activation('relu')(inception_4a_3x3)
inception_4a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_4a_3x3)
inception_4a_3x3 = Conv2D(192, (3, 3), strides=(1, 1), name='inception_4a_3x3_conv'+'2')(inception_4a_3x3)
inception_4a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_3x3_bn'+'2')(inception_4a_3x3)
inception_4a_3x3 = Activation('relu')(inception_4a_3x3)
inception_4a_5x5 = Conv2D(32, (1,1), strides=(1,1), name='inception_4a_5x5_conv1')(inception_3c) # inception 4a
inception_4a_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_5x5_bn1')(inception_4a_5x5) inception_4a_3x3 = Conv2D(
inception_4a_5x5 = Activation('relu')(inception_4a_5x5) 96, (1, 1), strides=(1, 1), name="inception_4a_3x3_conv" + "1"
inception_4a_5x5 = ZeroPadding2D(padding=(2,2))(inception_4a_5x5) )(inception_3c)
inception_4a_5x5 = Conv2D(64, (5,5), strides=(1,1), name='inception_4a_5x5_conv'+'2')(inception_4a_5x5) inception_4a_3x3 = BatchNormalization(
inception_4a_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_5x5_bn'+'2')(inception_4a_5x5) axis=3, epsilon=0.00001, name="inception_4a_3x3_bn" + "1"
inception_4a_5x5 = Activation('relu')(inception_4a_5x5) )(inception_4a_3x3)
inception_4a_3x3 = Activation("relu")(inception_4a_3x3)
inception_4a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_4a_3x3)
inception_4a_3x3 = Conv2D(
192, (3, 3), strides=(1, 1), name="inception_4a_3x3_conv" + "2"
)(inception_4a_3x3)
inception_4a_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4a_3x3_bn" + "2"
)(inception_4a_3x3)
inception_4a_3x3 = Activation("relu")(inception_4a_3x3)
inception_4a_pool = Lambda(lambda x: x**2, name='power2_4a')(inception_3c) inception_4a_5x5 = Conv2D(
inception_4a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(inception_4a_pool) 32, (1, 1), strides=(1, 1), name="inception_4a_5x5_conv1"
inception_4a_pool = Lambda(lambda x: x*9, name='mult9_4a')(inception_4a_pool) )(inception_3c)
inception_4a_pool = Lambda(lambda x: K.sqrt(x), name='sqrt_4a')(inception_4a_pool) inception_4a_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4a_5x5_bn1"
)(inception_4a_5x5)
inception_4a_5x5 = Activation("relu")(inception_4a_5x5)
inception_4a_5x5 = ZeroPadding2D(padding=(2, 2))(inception_4a_5x5)
inception_4a_5x5 = Conv2D(
64, (5, 5), strides=(1, 1), name="inception_4a_5x5_conv" + "2"
)(inception_4a_5x5)
inception_4a_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4a_5x5_bn" + "2"
)(inception_4a_5x5)
inception_4a_5x5 = Activation("relu")(inception_4a_5x5)
inception_4a_pool = Conv2D(128, (1,1), strides=(1,1), name='inception_4a_pool_conv'+'')(inception_4a_pool) inception_4a_pool = Lambda(lambda x: x ** 2, name="power2_4a")(inception_3c)
inception_4a_pool = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_pool_bn'+'')(inception_4a_pool) inception_4a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
inception_4a_pool = Activation('relu')(inception_4a_pool) inception_4a_pool
inception_4a_pool = ZeroPadding2D(padding=(2, 2))(inception_4a_pool) )
inception_4a_pool = Lambda(lambda x: x * 9, name="mult9_4a")(inception_4a_pool)
inception_4a_pool = Lambda(lambda x: K.sqrt(x), name="sqrt_4a")(inception_4a_pool)
inception_4a_1x1 = Conv2D(256, (1, 1), strides=(1, 1), name='inception_4a_1x1_conv'+'')(inception_3c) inception_4a_pool = Conv2D(
inception_4a_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_1x1_bn'+'')(inception_4a_1x1) 128, (1, 1), strides=(1, 1), name="inception_4a_pool_conv" + ""
inception_4a_1x1 = Activation('relu')(inception_4a_1x1) )(inception_4a_pool)
inception_4a_pool = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4a_pool_bn" + ""
)(inception_4a_pool)
inception_4a_pool = Activation("relu")(inception_4a_pool)
inception_4a_pool = ZeroPadding2D(padding=(2, 2))(inception_4a_pool)
inception_4a = concatenate([inception_4a_3x3, inception_4a_5x5, inception_4a_pool, inception_4a_1x1], axis=3) inception_4a_1x1 = Conv2D(
256, (1, 1), strides=(1, 1), name="inception_4a_1x1_conv" + ""
)(inception_3c)
inception_4a_1x1 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4a_1x1_bn" + ""
)(inception_4a_1x1)
inception_4a_1x1 = Activation("relu")(inception_4a_1x1)
#inception4e inception_4a = concatenate(
inception_4e_3x3 = Conv2D(160, (1,1), strides=(1,1), name='inception_4e_3x3_conv'+'1')(inception_4a) [inception_4a_3x3, inception_4a_5x5, inception_4a_pool, inception_4a_1x1],
inception_4e_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4e_3x3_bn'+'1')(inception_4e_3x3) axis=3,
inception_4e_3x3 = Activation('relu')(inception_4e_3x3) )
inception_4e_3x3 = ZeroPadding2D(padding=(1, 1))(inception_4e_3x3)
inception_4e_3x3 = Conv2D(256, (3,3), strides=(2,2), name='inception_4e_3x3_conv'+'2')(inception_4e_3x3)
inception_4e_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4e_3x3_bn'+'2')(inception_4e_3x3)
inception_4e_3x3 = Activation('relu')(inception_4e_3x3)
inception_4e_5x5 = Conv2D(64, (1,1), strides=(1,1), name='inception_4e_5x5_conv'+'1')(inception_4a) # inception4e
inception_4e_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4e_5x5_bn'+'1')(inception_4e_5x5) inception_4e_3x3 = Conv2D(
inception_4e_5x5 = Activation('relu')(inception_4e_5x5) 160, (1, 1), strides=(1, 1), name="inception_4e_3x3_conv" + "1"
inception_4e_5x5 = ZeroPadding2D(padding=(2, 2))(inception_4e_5x5) )(inception_4a)
inception_4e_5x5 = Conv2D(128, (5,5), strides=(2,2), name='inception_4e_5x5_conv'+'2')(inception_4e_5x5) inception_4e_3x3 = BatchNormalization(
inception_4e_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4e_5x5_bn'+'2')(inception_4e_5x5) axis=3, epsilon=0.00001, name="inception_4e_3x3_bn" + "1"
inception_4e_5x5 = Activation('relu')(inception_4e_5x5) )(inception_4e_3x3)
inception_4e_3x3 = Activation("relu")(inception_4e_3x3)
inception_4e_3x3 = ZeroPadding2D(padding=(1, 1))(inception_4e_3x3)
inception_4e_3x3 = Conv2D(
256, (3, 3), strides=(2, 2), name="inception_4e_3x3_conv" + "2"
)(inception_4e_3x3)
inception_4e_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4e_3x3_bn" + "2"
)(inception_4e_3x3)
inception_4e_3x3 = Activation("relu")(inception_4e_3x3)
inception_4e_pool = MaxPooling2D(pool_size=3, strides=2)(inception_4a) inception_4e_5x5 = Conv2D(
inception_4e_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_4e_pool) 64, (1, 1), strides=(1, 1), name="inception_4e_5x5_conv" + "1"
)(inception_4a)
inception_4e_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4e_5x5_bn" + "1"
)(inception_4e_5x5)
inception_4e_5x5 = Activation("relu")(inception_4e_5x5)
inception_4e_5x5 = ZeroPadding2D(padding=(2, 2))(inception_4e_5x5)
inception_4e_5x5 = Conv2D(
128, (5, 5), strides=(2, 2), name="inception_4e_5x5_conv" + "2"
)(inception_4e_5x5)
inception_4e_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4e_5x5_bn" + "2"
)(inception_4e_5x5)
inception_4e_5x5 = Activation("relu")(inception_4e_5x5)
inception_4e = concatenate([inception_4e_3x3, inception_4e_5x5, inception_4e_pool], axis=3) inception_4e_pool = MaxPooling2D(pool_size=3, strides=2)(inception_4a)
inception_4e_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_4e_pool)
#inception5a inception_4e = concatenate(
inception_5a_3x3 = Conv2D(96, (1,1), strides=(1,1), name='inception_5a_3x3_conv'+'1')(inception_4e) [inception_4e_3x3, inception_4e_5x5, inception_4e_pool], axis=3
inception_5a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5a_3x3_bn'+'1')(inception_5a_3x3) )
inception_5a_3x3 = Activation('relu')(inception_5a_3x3)
inception_5a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_5a_3x3)
inception_5a_3x3 = Conv2D(384, (3,3), strides=(1,1), name='inception_5a_3x3_conv'+'2')(inception_5a_3x3)
inception_5a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5a_3x3_bn'+'2')(inception_5a_3x3)
inception_5a_3x3 = Activation('relu')(inception_5a_3x3)
inception_5a_pool = Lambda(lambda x: x**2, name='power2_5a')(inception_4e) # inception5a
inception_5a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(inception_5a_pool) inception_5a_3x3 = Conv2D(
inception_5a_pool = Lambda(lambda x: x*9, name='mult9_5a')(inception_5a_pool) 96, (1, 1), strides=(1, 1), name="inception_5a_3x3_conv" + "1"
inception_5a_pool = Lambda(lambda x: K.sqrt(x), name='sqrt_5a')(inception_5a_pool) )(inception_4e)
inception_5a_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5a_3x3_bn" + "1"
)(inception_5a_3x3)
inception_5a_3x3 = Activation("relu")(inception_5a_3x3)
inception_5a_3x3 = ZeroPadding2D(padding=(1, 1))(inception_5a_3x3)
inception_5a_3x3 = Conv2D(
384, (3, 3), strides=(1, 1), name="inception_5a_3x3_conv" + "2"
)(inception_5a_3x3)
inception_5a_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5a_3x3_bn" + "2"
)(inception_5a_3x3)
inception_5a_3x3 = Activation("relu")(inception_5a_3x3)
inception_5a_pool = Conv2D(96, (1,1), strides=(1,1), name='inception_5a_pool_conv'+'')(inception_5a_pool) inception_5a_pool = Lambda(lambda x: x ** 2, name="power2_5a")(inception_4e)
inception_5a_pool = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5a_pool_bn'+'')(inception_5a_pool) inception_5a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
inception_5a_pool = Activation('relu')(inception_5a_pool) inception_5a_pool
inception_5a_pool = ZeroPadding2D(padding=(1,1))(inception_5a_pool) )
inception_5a_pool = Lambda(lambda x: x * 9, name="mult9_5a")(inception_5a_pool)
inception_5a_pool = Lambda(lambda x: K.sqrt(x), name="sqrt_5a")(inception_5a_pool)
inception_5a_1x1 = Conv2D(256, (1,1), strides=(1,1), name='inception_5a_1x1_conv'+'')(inception_4e) inception_5a_pool = Conv2D(
inception_5a_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5a_1x1_bn'+'')(inception_5a_1x1) 96, (1, 1), strides=(1, 1), name="inception_5a_pool_conv" + ""
inception_5a_1x1 = Activation('relu')(inception_5a_1x1) )(inception_5a_pool)
inception_5a_pool = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5a_pool_bn" + ""
)(inception_5a_pool)
inception_5a_pool = Activation("relu")(inception_5a_pool)
inception_5a_pool = ZeroPadding2D(padding=(1, 1))(inception_5a_pool)
inception_5a = concatenate([inception_5a_3x3, inception_5a_pool, inception_5a_1x1], axis=3) inception_5a_1x1 = Conv2D(
256, (1, 1), strides=(1, 1), name="inception_5a_1x1_conv" + ""
)(inception_4e)
inception_5a_1x1 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5a_1x1_bn" + ""
)(inception_5a_1x1)
inception_5a_1x1 = Activation("relu")(inception_5a_1x1)
#inception_5b inception_5a = concatenate(
inception_5b_3x3 = Conv2D(96, (1,1), strides=(1,1), name='inception_5b_3x3_conv'+'1')(inception_5a) [inception_5a_3x3, inception_5a_pool, inception_5a_1x1], axis=3
inception_5b_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5b_3x3_bn'+'1')(inception_5b_3x3) )
inception_5b_3x3 = Activation('relu')(inception_5b_3x3)
inception_5b_3x3 = ZeroPadding2D(padding=(1,1))(inception_5b_3x3)
inception_5b_3x3 = Conv2D(384, (3,3), strides=(1,1), name='inception_5b_3x3_conv'+'2')(inception_5b_3x3)
inception_5b_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5b_3x3_bn'+'2')(inception_5b_3x3)
inception_5b_3x3 = Activation('relu')(inception_5b_3x3)
inception_5b_pool = MaxPooling2D(pool_size=3, strides=2)(inception_5a) # inception_5b
inception_5b_3x3 = Conv2D(
96, (1, 1), strides=(1, 1), name="inception_5b_3x3_conv" + "1"
)(inception_5a)
inception_5b_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5b_3x3_bn" + "1"
)(inception_5b_3x3)
inception_5b_3x3 = Activation("relu")(inception_5b_3x3)
inception_5b_3x3 = ZeroPadding2D(padding=(1, 1))(inception_5b_3x3)
inception_5b_3x3 = Conv2D(
384, (3, 3), strides=(1, 1), name="inception_5b_3x3_conv" + "2"
)(inception_5b_3x3)
inception_5b_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5b_3x3_bn" + "2"
)(inception_5b_3x3)
inception_5b_3x3 = Activation("relu")(inception_5b_3x3)
inception_5b_pool = Conv2D(96, (1,1), strides=(1,1), name='inception_5b_pool_conv'+'')(inception_5b_pool) inception_5b_pool = MaxPooling2D(pool_size=3, strides=2)(inception_5a)
inception_5b_pool = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5b_pool_bn'+'')(inception_5b_pool)
inception_5b_pool = Activation('relu')(inception_5b_pool)
inception_5b_pool = ZeroPadding2D(padding=(1, 1))(inception_5b_pool) inception_5b_pool = Conv2D(
96, (1, 1), strides=(1, 1), name="inception_5b_pool_conv" + ""
)(inception_5b_pool)
inception_5b_pool = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5b_pool_bn" + ""
)(inception_5b_pool)
inception_5b_pool = Activation("relu")(inception_5b_pool)
inception_5b_1x1 = Conv2D(256, (1,1), strides=(1,1), name='inception_5b_1x1_conv'+'')(inception_5a) inception_5b_pool = ZeroPadding2D(padding=(1, 1))(inception_5b_pool)
inception_5b_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5b_1x1_bn'+'')(inception_5b_1x1)
inception_5b_1x1 = Activation('relu')(inception_5b_1x1)
inception_5b = concatenate([inception_5b_3x3, inception_5b_pool, inception_5b_1x1], axis=3) inception_5b_1x1 = Conv2D(
256, (1, 1), strides=(1, 1), name="inception_5b_1x1_conv" + ""
)(inception_5a)
inception_5b_1x1 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_5b_1x1_bn" + ""
)(inception_5b_1x1)
inception_5b_1x1 = Activation("relu")(inception_5b_1x1)
av_pool = AveragePooling2D(pool_size=(3, 3), strides=(1, 1))(inception_5b) inception_5b = concatenate(
reshape_layer = Flatten()(av_pool) [inception_5b_3x3, inception_5b_pool, inception_5b_1x1], axis=3
dense_layer = Dense(128, name='dense_layer')(reshape_layer) )
norm_layer = Lambda(lambda x: K.l2_normalize(x, axis=1), name='norm_layer')(dense_layer)
# Final Model av_pool = AveragePooling2D(pool_size=(3, 3), strides=(1, 1))(inception_5b)
model = Model(inputs=[myInput], outputs=norm_layer) reshape_layer = Flatten()(av_pool)
dense_layer = Dense(128, name="dense_layer")(reshape_layer)
norm_layer = Lambda(lambda x: K.l2_normalize(x, axis=1), name="norm_layer")(
dense_layer
)
#----------------------------------- # Final Model
return Model(inputs=[myInput], outputs=norm_layer)
home = str(Path.home())
if os.path.isfile(home+'/.deepface/weights/openface_weights.h5') != True: def loadModel(model_path=""):
print("openface_weights.h5 will be downloaded...") # -----------------------------------
if model_path:
assert Path(model_path).exists()
assert model_path.endswith(".h5")
url = 'https://drive.google.com/uc?id=1LSe1YCV1x-BfNnfb7DFZTNpv_Q9jITxn' else:
output = home+'/.deepface/weights/openface_weights.h5' home = Path.home().as_posix()
gdown.download(url, output, quiet=False) model_path = home + "/.deepface/weights/openface_weights.h5"
if not os.path.isfile(model_path):
print(f"openface_weights.h5 will be downloaded to {model_path}")
#----------------------------------- url = "https://drive.google.com/uc?id=1LSe1YCV1x-BfNnfb7DFZTNpv_Q9jITxn"
gdown.download(url, model_path, quiet=False)
model.load_weights(home+'/.deepface/weights/openface_weights.h5') # -----------------------------------
#----------------------------------- print(f"Loading model from {model_path}")
model = get_base_model()
model.load_weights(model_path)
return model # -----------------------------------
return model

138
deepface/basemodels/VGGFace.py
View File

@ -1,81 +1,99 @@
import os import os
from pathlib import Path from pathlib import Path
from keras.models import Model, Sequential from keras.models import Model, Sequential
from keras.layers import Input, Convolution2D, ZeroPadding2D, MaxPooling2D, Flatten, Dense, Dropout, Activation from keras.layers import (
Input,
Convolution2D,
ZeroPadding2D,
MaxPooling2D,
Flatten,
Dense,
Dropout,
Activation,
)
import gdown import gdown
#--------------------------------------- # ---------------------------------------
def baseModel():
model = Sequential()
model.add(ZeroPadding2D((1,1),input_shape=(224,224, 3)))
model.add(Convolution2D(64, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1))) def get_base_model():
model.add(Convolution2D(128, (3, 3), activation='relu')) model = Sequential()
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1), input_shape=(224, 224, 3)))
model.add(Convolution2D(128, (3, 3), activation='relu')) model.add(Convolution2D(64, (3, 3), activation="relu"))
model.add(MaxPooling2D((2,2), strides=(2,2))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(64, (3, 3), activation="relu"))
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(256, (3, 3), activation='relu')) model.add(Convolution2D(128, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(256, (3, 3), activation='relu')) model.add(Convolution2D(128, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(Convolution2D(256, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(MaxPooling2D((2,2), strides=(2,2))) model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1,1))) model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation='relu')) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(MaxPooling2D((2,2), strides=(2,2))) model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(Convolution2D(4096, (7, 7), activation='relu')) model.add(ZeroPadding2D((1, 1)))
model.add(Dropout(0.5)) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(Convolution2D(4096, (1, 1), activation='relu')) model.add(ZeroPadding2D((1, 1)))
model.add(Dropout(0.5)) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(Convolution2D(2622, (1, 1))) model.add(ZeroPadding2D((1, 1)))
model.add(Flatten()) model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(Activation('softmax')) model.add(MaxPooling2D((2, 2), strides=(2, 2)))
return model model.add(Convolution2D(4096, (7, 7), activation="relu"))
model.add(Dropout(0.5))
model.add(Convolution2D(4096, (1, 1), activation="relu"))
model.add(Dropout(0.5))
model.add(Convolution2D(2622, (1, 1)))
model.add(Flatten())
model.add(Activation("softmax"))
def loadModel(): return model
model = baseModel()
#----------------------------------- def loadModel(model_path=""):
"""
Args:
model_path: str
If provided, this path will be used to load the model from.
"""
if model_path:
assert Path(model_path).exists()
assert model_path.endswith(".h5")
else:
home = Path.home().as_posix()
model_path = os.path.join(home, ".deepface/weights/vgg_face_weights.h5")
if not os.path.isfile(model_path):
print(f"vgg_face_weights.h5 will be downloaded to {model_path}")
home = str(Path.home()) url = "https://drive.google.com/uc?id=1CPSeum3HpopfomUEK1gybeuIVoeJT_Eo"
gdown.download(url, model_path, quiet=False)
if os.path.isfile(home+'/.deepface/weights/vgg_face_weights.h5') != True: # -----------------------------------
print("vgg_face_weights.h5 will be downloaded...")
url = 'https://drive.google.com/uc?id=1CPSeum3HpopfomUEK1gybeuIVoeJT_Eo' print(f"Loading model from {model_path}")
output = home+'/.deepface/weights/vgg_face_weights.h5' model = get_base_model()
gdown.download(url, output, quiet=False) model.load_weights(model_path)
#----------------------------------- # -----------------------------------
model.load_weights(home+'/.deepface/weights/vgg_face_weights.h5') # TO-DO: why?
vgg_face_descriptor = Model(
inputs=model.layers[0].input, outputs=model.layers[-2].output
)
#----------------------------------- return vgg_face_descriptor
#TO-DO: why?
vgg_face_descriptor = Model(inputs=model.layers[0].input, outputs=model.layers[-2].output)
return vgg_face_descriptor

523
deepface/commons/functions.py
View File

@ -17,44 +17,52 @@ import subprocess
import tensorflow as tf import tensorflow as tf
import keras import keras
def loadBase64Img(uri): def loadBase64Img(uri):
encoded_data = uri.split(',')[1] encoded_data = uri.split(",")[1]
nparr = np.fromstring(base64.b64decode(encoded_data), np.uint8) nparr = np.fromstring(base64.b64decode(encoded_data), np.uint8)
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR) img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
return img return img
def distance(a, b): def distance(a, b):
x1 = a[0]; y1 = a[1] x1 = a[0]
x2 = b[0]; y2 = b[1] y1 = a[1]
x2 = b[0]
y2 = b[1]
return math.sqrt(((x2 - x1) * (x2 - x1)) + ((y2 - y1) * (y2 - y1)))
return math.sqrt(((x2 - x1) * (x2 - x1)) + ((y2 - y1) * (y2 - y1)))
def findFileHash(file): def findFileHash(file):
BLOCK_SIZE = 65536 # The size of each read from the file BLOCK_SIZE = 65536 # The size of each read from the file
file_hash = hashlib.sha256() # Create the hash object, can use something other than `.sha256()` if you wish file_hash = (
with open(file, 'rb') as f: # Open the file to read it's bytes hashlib.sha256()
fb = f.read(BLOCK_SIZE) # Read from the file. Take in the amount declared above ) # Create the hash object, can use something other than `.sha256()` if you wish
while len(fb) > 0: # While there is still data being read from the file with open(file, "rb") as f: # Open the file to read it's bytes
file_hash.update(fb) # Update the hash fb = f.read(BLOCK_SIZE) # Read from the file. Take in the amount declared above
fb = f.read(BLOCK_SIZE) # Read the next block from the file while len(fb) > 0: # While there is still data being read from the file
file_hash.update(fb) # Update the hash
fb = f.read(BLOCK_SIZE) # Read the next block from the file
return file_hash.hexdigest()
return file_hash.hexdigest()
def initializeFolder(): def initializeFolder():
home = str(Path.home()) home = str(Path.home())
if not os.path.exists(home+"/.deepface"): if not os.path.exists(home + "/.deepface"):
os.mkdir(home+"/.deepface") os.mkdir(home + "/.deepface")
print("Directory ",home,"/.deepface created") print("Directory ", home, "/.deepface created")
if not os.path.exists(home+"/.deepface/weights"): if not os.path.exists(home + "/.deepface/weights"):
os.mkdir(home+"/.deepface/weights") os.mkdir(home + "/.deepface/weights")
print("Directory ",home,"/.deepface/weights created") print("Directory ", home, "/.deepface/weights created")
#---------------------------------- # ----------------------------------
""" """
#avoid interrupted file download #avoid interrupted file download
weight_hashes = [ weight_hashes = [
@ -79,309 +87,346 @@ def initializeFolder():
print("hash violated for ", i[0],". It's going to be removed.") print("hash violated for ", i[0],". It's going to be removed.")
os.remove(weight_file) os.remove(weight_file)
""" """
#---------------------------------- # ----------------------------------
def findThreshold(model_name, distance_metric): def findThreshold(model_name, distance_metric):
threshold = 0.40 threshold = 0.40
if model_name == 'VGG-Face': if model_name == "VGG-Face":
if distance_metric == 'cosine': if distance_metric == "cosine":
threshold = 0.40 threshold = 0.40
elif distance_metric == 'euclidean': elif distance_metric == "euclidean":
threshold = 0.55 threshold = 0.55
elif distance_metric == 'euclidean_l2': elif distance_metric == "euclidean_l2":
threshold = 0.75 threshold = 0.75
elif model_name == 'OpenFace': elif model_name == "OpenFace":
if distance_metric == 'cosine': if distance_metric == "cosine":
threshold = 0.10 threshold = 0.10
elif distance_metric == 'euclidean': elif distance_metric == "euclidean":
threshold = 0.55 threshold = 0.55
elif distance_metric == 'euclidean_l2': elif distance_metric == "euclidean_l2":
threshold = 0.55 threshold = 0.55
elif model_name == 'Facenet': elif model_name == "Facenet":
if distance_metric == 'cosine': if distance_metric == "cosine":
threshold = 0.40 threshold = 0.40
elif distance_metric == 'euclidean': elif distance_metric == "euclidean":
threshold = 10 threshold = 10
elif distance_metric == 'euclidean_l2': elif distance_metric == "euclidean_l2":
threshold = 0.80 threshold = 0.80
elif model_name == 'DeepFace': elif model_name == "DeepFace":
if distance_metric == 'cosine': if distance_metric == "cosine":
threshold = 0.23 threshold = 0.23
elif distance_metric == 'euclidean': elif distance_metric == "euclidean":
threshold = 64 threshold = 64
elif distance_metric == 'euclidean_l2': elif distance_metric == "euclidean_l2":
threshold = 0.64 threshold = 0.64
return threshold
return threshold
def get_opencv_path(): def get_opencv_path():
opencv_home = cv2.__file__ opencv_home = cv2.__file__
folders = opencv_home.split(os.path.sep)[0:-1] folders = opencv_home.split(os.path.sep)[0:-1]
path = folders[0] path = folders[0]
for folder in folders[1:]: for folder in folders[1:]:
path = path + "/" + folder path = path + "/" + folder
face_detector_path = path+"/data/haarcascade_frontalface_default.xml" face_detector_path = path + "/data/haarcascade_frontalface_default.xml"
eye_detector_path = path+"/data/haarcascade_eye.xml" eye_detector_path = path + "/data/haarcascade_eye.xml"
if os.path.isfile(face_detector_path) != True: if os.path.isfile(face_detector_path) != True:
raise ValueError("Confirm that opencv is installed on your environment! Expected path ",face_detector_path," violated.") raise ValueError(
"Confirm that opencv is installed on your environment! Expected path ",
face_detector_path,
" violated.",
)
return path+"/data/" return path + "/data/"
def detectFace(img, target_size=(224, 224), grayscale = False, enforce_detection = True):
img_path = "" def detectFace(img, target_size=(224, 224), grayscale=False, enforce_detection=True):
#----------------------- img_path = ""
exact_image = False # -----------------------
if type(img).__module__ == np.__name__:
exact_image = True
base64_img = False exact_image = False
if len(img) > 11 and img[0:11] == "data:image/": if type(img).__module__ == np.__name__:
base64_img = True exact_image = True
#----------------------- base64_img = False
if len(img) > 11 and img[0:11] == "data:image/":
base64_img = True
opencv_path = get_opencv_path() # -----------------------
face_detector_path = opencv_path+"haarcascade_frontalface_default.xml"
eye_detector_path = opencv_path+"haarcascade_eye.xml"
if os.path.isfile(face_detector_path) != True: opencv_path = get_opencv_path()
raise ValueError("Confirm that opencv is installed on your environment! Expected path ",face_detector_path," violated.") face_detector_path = opencv_path + "haarcascade_frontalface_default.xml"
eye_detector_path = opencv_path + "haarcascade_eye.xml"
#-------------------------------- if os.path.isfile(face_detector_path) != True:
raise ValueError(
"Confirm that opencv is installed on your environment! Expected path ",
face_detector_path,
" violated.",
)
face_detector = cv2.CascadeClassifier(face_detector_path) # --------------------------------
eye_detector = cv2.CascadeClassifier(eye_detector_path)
if base64_img == True: face_detector = cv2.CascadeClassifier(face_detector_path)
img = loadBase64Img(img) eye_detector = cv2.CascadeClassifier(eye_detector_path)
elif exact_image != True: #image path passed as input if base64_img == True:
img = loadBase64Img(img)
if os.path.isfile(img) != True: elif exact_image != True: # image path passed as input
raise ValueError("Confirm that ",img," exists")
img = cv2.imread(img) if os.path.isfile(img) != True:
raise ValueError("Confirm that ", img, " exists")
img_raw = img.copy() img = cv2.imread(img)
#-------------------------------- img_raw = img.copy()
faces = face_detector.detectMultiScale(img, 1.3, 5) # --------------------------------
#print("found faces in ",image_path," is ",len(faces)) faces = face_detector.detectMultiScale(img, 1.3, 5)
if len(faces) > 0: # print("found faces in ",image_path," is ",len(faces))
x,y,w,h = faces[0]
detected_face = img[int(y):int(y+h), int(x):int(x+w)]
detected_face_gray = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
#--------------------------- if len(faces) > 0:
#face alignment x, y, w, h = faces[0]
detected_face = img[int(y) : int(y + h), int(x) : int(x + w)]
detected_face_gray = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
eyes = eye_detector.detectMultiScale(detected_face_gray) # ---------------------------
# face alignment
if len(eyes) >= 2: eyes = eye_detector.detectMultiScale(detected_face_gray)
#find the largest 2 eye
base_eyes = eyes[:, 2]
items = [] if len(eyes) >= 2:
for i in range(0, len(base_eyes)): # find the largest 2 eye
item = (base_eyes[i], i) base_eyes = eyes[:, 2]
items.append(item)
df = pd.DataFrame(items, columns = ["length", "idx"]).sort_values(by=['length'], ascending=False) items = []
for i in range(0, len(base_eyes)):
item = (base_eyes[i], i)
items.append(item)
eyes = eyes[df.idx.values[0:2]] df = pd.DataFrame(items, columns=["length", "idx"]).sort_values(
by=["length"], ascending=False
)
#----------------------- eyes = eyes[df.idx.values[0:2]]
#decide left and right eye
eye_1 = eyes[0]; eye_2 = eyes[1] # -----------------------
# decide left and right eye
if eye_1[0] < eye_2[0]: eye_1 = eyes[0]
left_eye = eye_1 eye_2 = eyes[1]
right_eye = eye_2
else:
left_eye = eye_2
right_eye = eye_1
#----------------------- if eye_1[0] < eye_2[0]:
#find center of eyes left_eye = eye_1
right_eye = eye_2
else:
left_eye = eye_2
right_eye = eye_1
left_eye_center = (int(left_eye[0] + (left_eye[2] / 2)), int(left_eye[1] + (left_eye[3] / 2))) # -----------------------
left_eye_x = left_eye_center[0]; left_eye_y = left_eye_center[1] # find center of eyes
right_eye_center = (int(right_eye[0] + (right_eye[2]/2)), int(right_eye[1] + (right_eye[3]/2))) left_eye_center = (
right_eye_x = right_eye_center[0]; right_eye_y = right_eye_center[1] int(left_eye[0] + (left_eye[2] / 2)),
int(left_eye[1] + (left_eye[3] / 2)),
)
left_eye_x = left_eye_center[0]
left_eye_y = left_eye_center[1]
#----------------------- right_eye_center = (
#find rotation direction int(right_eye[0] + (right_eye[2] / 2)),
int(right_eye[1] + (right_eye[3] / 2)),
)
right_eye_x = right_eye_center[0]
right_eye_y = right_eye_center[1]
if left_eye_y > right_eye_y: # -----------------------
point_3rd = (right_eye_x, left_eye_y) # find rotation direction
direction = -1 #rotate same direction to clock
else:
point_3rd = (left_eye_x, right_eye_y)
direction = 1 #rotate inverse direction of clock
#----------------------- if left_eye_y > right_eye_y:
#find length of triangle edges point_3rd = (right_eye_x, left_eye_y)
direction = -1 # rotate same direction to clock
else:
point_3rd = (left_eye_x, right_eye_y)
direction = 1 # rotate inverse direction of clock
a = distance(left_eye_center, point_3rd) # -----------------------
b = distance(right_eye_center, point_3rd) # find length of triangle edges
c = distance(right_eye_center, left_eye_center)
#----------------------- a = distance(left_eye_center, point_3rd)
#apply cosine rule b = distance(right_eye_center, point_3rd)
c = distance(right_eye_center, left_eye_center)
cos_a = (b*b + c*c - a*a)/(2*b*c) # -----------------------
angle = np.arccos(cos_a) #angle in radian # apply cosine rule
angle = (angle * 180) / math.pi #radian to degree
#----------------------- cos_a = (b * b + c * c - a * a) / (2 * b * c)
#rotate base image angle = np.arccos(cos_a) # angle in radian
angle = (angle * 180) / math.pi # radian to degree
if direction == -1: # -----------------------
angle = 90 - angle # rotate base image
img = Image.fromarray(img_raw) if direction == -1:
img = np.array(img.rotate(direction * angle)) angle = 90 - angle
#you recover the base image and face detection disappeared. apply again. img = Image.fromarray(img_raw)
faces = face_detector.detectMultiScale(img, 1.3, 5) img = np.array(img.rotate(direction * angle))
if len(faces) > 0:
x,y,w,h = faces[0]
detected_face = img[int(y):int(y+h), int(x):int(x+w)]
#----------------------- # you recover the base image and face detection disappeared. apply again.
faces = face_detector.detectMultiScale(img, 1.3, 5)
if len(faces) > 0:
x, y, w, h = faces[0]
detected_face = img[int(y) : int(y + h), int(x) : int(x + w)]
#face alignment block end # -----------------------
#---------------------------
#face alignment block needs colorful images. that's why, converting to gray scale logic moved to here. # face alignment block end
if grayscale == True: # ---------------------------
detected_face = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
detected_face = cv2.resize(detected_face, target_size) # face alignment block needs colorful images. that's why, converting to gray scale logic moved to here.
if grayscale == True:
detected_face = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
img_pixels = image.img_to_array(detected_face) detected_face = cv2.resize(detected_face, target_size)
img_pixels = np.expand_dims(img_pixels, axis = 0)
#normalize input in [0, 1] img_pixels = image.img_to_array(detected_face)
img_pixels /= 255 img_pixels = np.expand_dims(img_pixels, axis=0)
return img_pixels # normalize input in [0, 1]
img_pixels /= 255
else: return img_pixels
if (exact_image == True) or (enforce_detection != True): else:
if grayscale == True: if (exact_image == True) or (enforce_detection != True):
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
if grayscale == True:
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img = cv2.resize(img, target_size)
img_pixels = image.img_to_array(img)
img_pixels = np.expand_dims(img_pixels, axis=0)
img_pixels /= 255
return img_pixels
else:
raise ValueError(
"Face could not be detected. Please confirm that the picture is a face photo or consider to set enforce_detection param to False."
)
img = cv2.resize(img, target_size)
img_pixels = image.img_to_array(img)
img_pixels = np.expand_dims(img_pixels, axis = 0)
img_pixels /= 255
return img_pixels
else:
raise ValueError("Face could not be detected. Please confirm that the picture is a face photo or consider to set enforce_detection param to False.")
def allocateMemory(): def allocateMemory():
#find allocated memories # find allocated memories
gpu_indexes = [] gpu_indexes = []
memory_usage_percentages = []; available_memories = []; total_memories = []; utilizations = [] memory_usage_percentages = []
power_usages = []; power_capacities = [] available_memories = []
total_memories = []
utilizations = []
power_usages = []
power_capacities = []
try: try:
result = subprocess.check_output(['nvidia-smi']) result = subprocess.check_output(["nvidia-smi"])
dashboard = result.decode("utf-8").split("=|") dashboard = result.decode("utf-8").split("=|")
dashboard = dashboard[1].split("\n") dashboard = dashboard[1].split("\n")
gpu_idx = 0 gpu_idx = 0
for line in dashboard: for line in dashboard:
if ("MiB" in line): if "MiB" in line:
power_info = line.split("|")[1] power_info = line.split("|")[1]
power_capacity = int(power_info.split("/")[-1].replace("W", "")) power_capacity = int(power_info.split("/")[-1].replace("W", ""))
power_usage = int((power_info.split("/")[-2]).strip().split(" ")[-1].replace("W", "")) power_usage = int(
(power_info.split("/")[-2]).strip().split(" ")[-1].replace("W", "")
)
power_usages.append(power_usage) power_usages.append(power_usage)
power_capacities.append(power_capacity) power_capacities.append(power_capacity)
#---------------------------- # ----------------------------
memory_info = line.split("|")[2].replace("MiB","").split("/") memory_info = line.split("|")[2].replace("MiB", "").split("/")
utilization_info = int(line.split("|")[3].split("%")[0]) utilization_info = int(line.split("|")[3].split("%")[0])
allocated = int(memory_info[0]) allocated = int(memory_info[0])
total_memory = int(memory_info[1]) total_memory = int(memory_info[1])
available_memory = total_memory - allocated available_memory = total_memory - allocated
total_memories.append(total_memory) total_memories.append(total_memory)
available_memories.append(available_memory) available_memories.append(available_memory)
memory_usage_percentages.append(round(100*int(allocated)/int(total_memory), 4)) memory_usage_percentages.append(
utilizations.append(utilization_info) round(100 * int(allocated) / int(total_memory), 4)
gpu_indexes.append(gpu_idx) )
utilizations.append(utilization_info)
gpu_indexes.append(gpu_idx)
gpu_idx = gpu_idx + 1 gpu_idx = gpu_idx + 1
gpu_count = gpu_idx * 1 gpu_count = gpu_idx * 1
except Exception as err: except Exception as err:
gpu_count = 0 gpu_count = 0
#print(str(err)) # print(str(err))
#------------------------------ # ------------------------------
df = pd.DataFrame(gpu_indexes, columns = ["gpu_index"]) df = pd.DataFrame(gpu_indexes, columns=["gpu_index"])
df["total_memories_in_mb"] = total_memories df["total_memories_in_mb"] = total_memories
df["available_memories_in_mb"] = available_memories df["available_memories_in_mb"] = available_memories
df["memory_usage_percentage"] = memory_usage_percentages df["memory_usage_percentage"] = memory_usage_percentages
df["utilizations"] = utilizations df["utilizations"] = utilizations
df["power_usages_in_watts"] = power_usages df["power_usages_in_watts"] = power_usages
df["power_capacities_in_watts"] = power_capacities df["power_capacities_in_watts"] = power_capacities
df = df.sort_values(by = ["available_memories_in_mb"], ascending = False).reset_index(drop = True) df = df.sort_values(by=["available_memories_in_mb"], ascending=False).reset_index(
drop=True
)
#------------------------------ # ------------------------------
required_memory = 10000 #All deepface models require 9016 MiB required_memory = 10000 # All deepface models require 9016 MiB
if df.shape[0] > 0: #has gpu if df.shape[0] > 0: # has gpu
if df.iloc[0].available_memories_in_mb > required_memory: if df.iloc[0].available_memories_in_mb > required_memory:
my_gpu = str(int(df.iloc[0].gpu_index)) my_gpu = str(int(df.iloc[0].gpu_index))
os.environ["CUDA_VISIBLE_DEVICES"] = my_gpu os.environ["CUDA_VISIBLE_DEVICES"] = my_gpu
#------------------------------ # ------------------------------
#tf allocates all memory by default # tf allocates all memory by default
#this block avoids greedy approach # this block avoids greedy approach
config = tf.ConfigProto() config = tf.ConfigProto()
config.gpu_options.allow_growth = True config.gpu_options.allow_growth = True
session = tf.Session(config=config) session = tf.Session(config=config)
keras.backend.set_session(session) keras.backend.set_session(session)
print("DeepFace will run on GPU (gpu_", my_gpu,")") print("DeepFace will run on GPU (gpu_", my_gpu, ")")
else: else:
#this case has gpu but no enough memory to allocate # this case has gpu but no enough memory to allocate
os.environ["CUDA_VISIBLE_DEVICES"] = "" #run it on cpu os.environ["CUDA_VISIBLE_DEVICES"] = "" # run it on cpu
print("Even though the system has GPUs, there is no enough space in memory to allocate.") print(
print("DeepFace will run on CPU") "Even though the system has GPUs, there is no enough space in memory to allocate."
else: )
print("DeepFace will run on CPU") print("DeepFace will run on CPU")
else:
print("DeepFace will run on CPU")
#------------------------------ # ------------------------------