github/deepface
1
1
Fork 0
mirror of https://github.com/serengil/deepface.git synced 2025-06-09 12:57:08 +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

2
.gitignore vendored
View File

@ -13,4 +13,4 @@ deepface/__pycache__/*
deepface/commons/__pycache__/*
deepface/basemodels/__pycache__/*
deepface/extendedmodels/__pycache__/*
deepface/subsidiarymodels/__pycache__/*
deepface/subsidiarymodels/__pycache__/*

646
deepface/DeepFace.py
View File

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

2732
deepface/basemodels/Facenet.py
View File

File diff suppressed because it is too large Load Diff

103
deepface/basemodels/FbDeepFace.py
View File

@ -3,44 +3,71 @@ from pathlib import Path
import gdown
import keras
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
#-------------------------------------
# -------------------------------------
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'))
#---------------------------------
home = str(Path.home())
if os.path.isfile(home+'/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5') != True:
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'
output = home+'/.deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5.zip'
gdown.download(url, output, quiet=False)
#unzip VGGFace2_DeepFace_weights_val-0.9034.h5.zip
with zipfile.ZipFile(output, 'r') as zip_ref:
zip_ref.extractall(home+'/.deepface/weights/')
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)
return deepface_model
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
def loadModel(model_path=""):
# ---------------------------------
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"
zip_path = os.path.join(
home, ".deepface/weights/VGGFace2_DeepFace_weights_val-0.9034.h5.zip"
)
gdown.download(url, zip_path, 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/"))
print(f"Loading model from {model_path}")
base_model = get_base_model()
base_model.load_weights(model_path)
# drop F8 and D0. F7 is the representation layer.
deepface_model = Model(
inputs=base_model.layers[0].input, outputs=base_model.layers[-3].output
)
return deepface_model

551
deepface/basemodels/OpenFace.py
View File

@ -13,238 +13,397 @@ from keras.layers.normalization import BatchNormalization
from keras.models import load_model
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)
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1')(x)
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)
def get_base_model():
myInput = Input(shape=(96, 96, 3))
# Inception3a
inception_3a_3x3 = Conv2D(96, (1, 1), name='inception_3a_3x3_conv1')(x)
inception_3a_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_3x3_bn1')(inception_3a_3x3)
inception_3a_3x3 = Activation('relu')(inception_3a_3x3)
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)
x = ZeroPadding2D(padding=(3, 3), input_shape=(96, 96, 3))(myInput)
x = Conv2D(64, (7, 7), strides=(2, 2), name="conv1")(x)
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)
inception_3a_5x5 = Conv2D(16, (1, 1), name='inception_3a_5x5_conv1')(x)
inception_3a_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_5x5_bn1')(inception_3a_5x5)
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)
# Inception3a
inception_3a_3x3 = Conv2D(96, (1, 1), name="inception_3a_3x3_conv1")(x)
inception_3a_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_3x3_bn1"
)(inception_3a_3x3)
inception_3a_3x3 = Activation("relu")(inception_3a_3x3)
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_pool = Conv2D(32, (1, 1), name='inception_3a_pool_conv')(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_5x5 = Conv2D(16, (1, 1), name="inception_3a_5x5_conv1")(x)
inception_3a_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3a_5x5_bn1"
)(inception_3a_5x5)
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_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3a_1x1_bn')(inception_3a_1x1)
inception_3a_1x1 = Activation('relu')(inception_3a_1x1)
inception_3a_pool = MaxPooling2D(pool_size=3, strides=2)(x)
inception_3a_pool = Conv2D(32, (1, 1), name="inception_3a_pool_conv")(
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_3b_3x3 = Conv2D(96, (1, 1), name='inception_3b_3x3_conv1')(inception_3a)
inception_3b_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_3x3_bn1')(inception_3b_3x3)
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_3a = concatenate(
[inception_3a_3x3, inception_3a_5x5, inception_3a_pool, inception_3a_1x1],
axis=3,
)
inception_3b_5x5 = Conv2D(32, (1, 1), name='inception_3b_5x5_conv1')(inception_3a)
inception_3b_5x5 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_5x5_bn1')(inception_3b_5x5)
inception_3b_5x5 = Activation('relu')(inception_3b_5x5)
inception_3b_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3b_5x5)
inception_3b_5x5 = Conv2D(64, (5, 5), name='inception_3b_5x5_conv2')(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)
# Inception3b
inception_3b_3x3 = Conv2D(96, (1, 1), name="inception_3b_3x3_conv1")(inception_3a)
inception_3b_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_3x3_bn1"
)(inception_3b_3x3)
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_pool = Lambda(lambda x: x**2, name='power2_3b')(inception_3a)
inception_3b_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(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_5x5 = Conv2D(32, (1, 1), name="inception_3b_5x5_conv1")(inception_3a)
inception_3b_5x5 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3b_5x5_bn1"
)(inception_3b_5x5)
inception_3b_5x5 = Activation("relu")(inception_3b_5x5)
inception_3b_5x5 = ZeroPadding2D(padding=(2, 2))(inception_3b_5x5)
inception_3b_5x5 = Conv2D(64, (5, 5), name="inception_3b_5x5_conv2")(
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_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3b_1x1_bn')(inception_3b_1x1)
inception_3b_1x1 = Activation('relu')(inception_3b_1x1)
inception_3b_pool = Lambda(lambda x: x ** 2, name="power2_3b")(inception_3a)
inception_3b_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
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_3c_3x3 = Conv2D(128, (1, 1), strides=(1, 1), name='inception_3c_3x3_conv1')(inception_3b)
inception_3c_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_3c_3x3_bn1')(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_3b = concatenate(
[inception_3b_3x3, inception_3b_5x5, inception_3b_pool, inception_3b_1x1],
axis=3,
)
inception_3c_5x5 = Conv2D(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)
# Inception3c
inception_3c_3x3 = Conv2D(
128, (1, 1), strides=(1, 1), name="inception_3c_3x3_conv1"
)(inception_3b)
inception_3c_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_3c_3x3_bn1"
)(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_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_3c_pool)
inception_3c_5x5 = Conv2D(
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_4a_3x3 = Conv2D(96, (1, 1), strides=(1, 1), name='inception_4a_3x3_conv'+'1')(inception_3c)
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_3c = concatenate(
[inception_3c_3x3, inception_3c_5x5, inception_3c_pool], axis=3
)
inception_4a_5x5 = Conv2D(32, (1,1), strides=(1,1), name='inception_4a_5x5_conv1')(inception_3c)
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
inception_4a_3x3 = Conv2D(
96, (1, 1), strides=(1, 1), name="inception_4a_3x3_conv" + "1"
)(inception_3c)
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_pool = Lambda(lambda x: x**2, name='power2_4a')(inception_3c)
inception_4a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(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_5x5 = Conv2D(
32, (1, 1), strides=(1, 1), name="inception_4a_5x5_conv1"
)(inception_3c)
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 = 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_pool = Lambda(lambda x: x ** 2, name="power2_4a")(inception_3c)
inception_4a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
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_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4a_1x1_bn'+'')(inception_4a_1x1)
inception_4a_1x1 = Activation('relu')(inception_4a_1x1)
inception_4a_pool = Conv2D(
128, (1, 1), strides=(1, 1), name="inception_4a_pool_conv" + ""
)(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_4e_3x3 = Conv2D(160, (1,1), strides=(1,1), name='inception_4e_3x3_conv'+'1')(inception_4a)
inception_4e_3x3 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_4e_3x3_bn'+'1')(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_4a = concatenate(
[inception_4a_3x3, inception_4a_5x5, inception_4a_pool, inception_4a_1x1],
axis=3,
)
inception_4e_5x5 = Conv2D(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)
# inception4e
inception_4e_3x3 = Conv2D(
160, (1, 1), strides=(1, 1), name="inception_4e_3x3_conv" + "1"
)(inception_4a)
inception_4e_3x3 = BatchNormalization(
axis=3, epsilon=0.00001, name="inception_4e_3x3_bn" + "1"
)(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_pool = ZeroPadding2D(padding=((0, 1), (0, 1)))(inception_4e_pool)
inception_4e_5x5 = Conv2D(
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_5a_3x3 = Conv2D(96, (1,1), strides=(1,1), name='inception_5a_3x3_conv'+'1')(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_4e = concatenate(
[inception_4e_3x3, inception_4e_5x5, inception_4e_pool], axis=3
)
inception_5a_pool = Lambda(lambda x: x**2, name='power2_5a')(inception_4e)
inception_5a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(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)
# inception5a
inception_5a_3x3 = Conv2D(
96, (1, 1), strides=(1, 1), name="inception_5a_3x3_conv" + "1"
)(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 = 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_pool = Lambda(lambda x: x ** 2, name="power2_5a")(inception_4e)
inception_5a_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3))(
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_1x1 = BatchNormalization(axis=3, epsilon=0.00001, name='inception_5a_1x1_bn'+'')(inception_5a_1x1)
inception_5a_1x1 = Activation('relu')(inception_5a_1x1)
inception_5a_pool = Conv2D(
96, (1, 1), strides=(1, 1), name="inception_5a_pool_conv" + ""
)(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_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_5a = concatenate(
[inception_5a_3x3, inception_5a_pool, inception_5a_1x1], axis=3
)
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 = 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 = MaxPooling2D(pool_size=3, strides=2)(inception_5a)
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_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_pool = ZeroPadding2D(padding=(1, 1))(inception_5b_pool)
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)
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)
inception_5b = concatenate(
[inception_5b_3x3, inception_5b_pool, inception_5b_1x1], axis=3
)
# Final Model
model = Model(inputs=[myInput], outputs=norm_layer)
#-----------------------------------
home = str(Path.home())
if os.path.isfile(home+'/.deepface/weights/openface_weights.h5') != True:
print("openface_weights.h5 will be downloaded...")
url = 'https://drive.google.com/uc?id=1LSe1YCV1x-BfNnfb7DFZTNpv_Q9jITxn'
output = home+'/.deepface/weights/openface_weights.h5'
gdown.download(url, output, quiet=False)
#-----------------------------------
model.load_weights(home+'/.deepface/weights/openface_weights.h5')
#-----------------------------------
return model
av_pool = AveragePooling2D(pool_size=(3, 3), strides=(1, 1))(inception_5b)
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)
def loadModel(model_path=""):
# -----------------------------------
if model_path:
assert Path(model_path).exists()
assert model_path.endswith(".h5")
else:
home = Path.home().as_posix()
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)
# -----------------------------------
print(f"Loading model from {model_path}")
model = get_base_model()
model.load_weights(model_path)
# -----------------------------------
return model

156
deepface/basemodels/VGGFace.py
View File

@ -1,81 +1,99 @@
import os
from pathlib import Path
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
#---------------------------------------
# ---------------------------------------
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)))
model.add(Convolution2D(128, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(128, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
def get_base_model():
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)))
model.add(Convolution2D(256, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(128, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(128, (3, 3), activation="relu"))
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(256, (3, 3), activation="relu"))
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, (3, 3), activation='relu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
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'))
return model
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(ZeroPadding2D((1, 1)))
model.add(Convolution2D(512, (3, 3), activation="relu"))
model.add(MaxPooling2D((2, 2), strides=(2, 2)))
def loadModel():
model = baseModel()
#-----------------------------------
home = str(Path.home())
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'
output = home+'/.deepface/weights/vgg_face_weights.h5'
gdown.download(url, output, quiet=False)
#-----------------------------------
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
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"))
return model
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}")
url = "https://drive.google.com/uc?id=1CPSeum3HpopfomUEK1gybeuIVoeJT_Eo"
gdown.download(url, model_path, quiet=False)
# -----------------------------------
print(f"Loading model from {model_path}")
model = get_base_model()
model.load_weights(model_path)
# -----------------------------------
# TO-DO: why?
vgg_face_descriptor = Model(
inputs=model.layers[0].input, outputs=model.layers[-2].output
)
return vgg_face_descriptor

705
deepface/commons/functions.py
View File

@ -17,46 +17,54 @@ import subprocess
import tensorflow as tf
import keras
def loadBase64Img(uri):
encoded_data = uri.split(',')[1]
nparr = np.fromstring(base64.b64decode(encoded_data), np.uint8)
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
return img
encoded_data = uri.split(",")[1]
nparr = np.fromstring(base64.b64decode(encoded_data), np.uint8)
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
return img
def distance(a, b):
x1 = a[0]; y1 = a[1]
x2 = b[0]; y2 = b[1]
return math.sqrt(((x2 - x1) * (x2 - x1)) + ((y2 - y1) * (y2 - y1)))
x1 = a[0]
y1 = a[1]
x2 = b[0]
y2 = b[1]
return math.sqrt(((x2 - x1) * (x2 - x1)) + ((y2 - y1) * (y2 - y1)))
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
with open(file, "rb") as f: # Open the file to read it's bytes
fb = f.read(BLOCK_SIZE) # Read from the file. Take in the amount declared above
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()
file_hash = hashlib.sha256() # Create the hash object, can use something other than `.sha256()` if you wish
with open(file, 'rb') as f: # Open the file to read it's bytes
fb = f.read(BLOCK_SIZE) # Read from the file. Take in the amount declared above
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()
def initializeFolder():
home = str(Path.home())
if not os.path.exists(home+"/.deepface"):
os.mkdir(home+"/.deepface")
print("Directory ",home,"/.deepface created")
if not os.path.exists(home+"/.deepface/weights"):
os.mkdir(home+"/.deepface/weights")
print("Directory ",home,"/.deepface/weights created")
#----------------------------------
"""
home = str(Path.home())
if not os.path.exists(home + "/.deepface"):
os.mkdir(home + "/.deepface")
print("Directory ", home, "/.deepface created")
if not os.path.exists(home + "/.deepface/weights"):
os.mkdir(home + "/.deepface/weights")
print("Directory ", home, "/.deepface/weights created")
# ----------------------------------
"""
#avoid interrupted file download
weight_hashes = [
['age_model_weights.h5', '0aeff75734bfe794113756d2bfd0ac823d51e9422c8961125b570871d3c2b114']
, ['facenet_weights.h5', '90659cc97bfda5999120f95d8e122f4d262cca11715a21e59ba024bcce816d5c']
@ -66,12 +74,12 @@ def initializeFolder():
, ['race_model_single_batch.h5', 'eb22b28b1f6dfce65b64040af4e86003a5edccb169a1a338470dde270b6f5e54']
, ['vgg_face_weights.h5', '759266b9614d0fd5d65b97bf716818b746cc77ab5944c7bffc937c6ba9455d8c']
]
for i in weight_hashes:
weight_file = home+"/.deepface/weights/"+i[0]
expected_hash = i[1]
#check file exits
if os.path.isfile(weight_file) == True:
current_hash = findFileHash(weight_file)
@ -79,309 +87,346 @@ def initializeFolder():
print("hash violated for ", i[0],". It's going to be removed.")
os.remove(weight_file)
"""
#----------------------------------
# ----------------------------------
def findThreshold(model_name, distance_metric):
threshold = 0.40
if model_name == 'VGG-Face':
if distance_metric == 'cosine':
threshold = 0.40
elif distance_metric == 'euclidean':
threshold = 0.55
elif distance_metric == 'euclidean_l2':
threshold = 0.75
elif model_name == 'OpenFace':
if distance_metric == 'cosine':
threshold = 0.10
elif distance_metric == 'euclidean':
threshold = 0.55
elif distance_metric == 'euclidean_l2':
threshold = 0.55
elif model_name == 'Facenet':
if distance_metric == 'cosine':
threshold = 0.40
elif distance_metric == 'euclidean':
threshold = 10
elif distance_metric == 'euclidean_l2':
threshold = 0.80
elif model_name == 'DeepFace':
if distance_metric == 'cosine':
threshold = 0.23
elif distance_metric == 'euclidean':
threshold = 64
elif distance_metric == 'euclidean_l2':
threshold = 0.64
return threshold
threshold = 0.40
if model_name == "VGG-Face":
if distance_metric == "cosine":
threshold = 0.40
elif distance_metric == "euclidean":
threshold = 0.55
elif distance_metric == "euclidean_l2":
threshold = 0.75
elif model_name == "OpenFace":
if distance_metric == "cosine":
threshold = 0.10
elif distance_metric == "euclidean":
threshold = 0.55
elif distance_metric == "euclidean_l2":
threshold = 0.55
elif model_name == "Facenet":
if distance_metric == "cosine":
threshold = 0.40
elif distance_metric == "euclidean":
threshold = 10
elif distance_metric == "euclidean_l2":
threshold = 0.80
elif model_name == "DeepFace":
if distance_metric == "cosine":
threshold = 0.23
elif distance_metric == "euclidean":
threshold = 64
elif distance_metric == "euclidean_l2":
threshold = 0.64
return threshold
def get_opencv_path():
opencv_home = cv2.__file__
folders = opencv_home.split(os.path.sep)[0:-1]
path = folders[0]
for folder in folders[1:]:
path = path + "/" + folder
opencv_home = cv2.__file__
folders = opencv_home.split(os.path.sep)[0:-1]
face_detector_path = path+"/data/haarcascade_frontalface_default.xml"
eye_detector_path = path+"/data/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.")
return path+"/data/"
path = folders[0]
for folder in folders[1:]:
path = path + "/" + folder
face_detector_path = path + "/data/haarcascade_frontalface_default.xml"
eye_detector_path = path + "/data/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.",
)
return path + "/data/"
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
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:
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:
img = loadBase64Img(img)
elif exact_image != True: # image path passed as input
if os.path.isfile(img) != True:
raise ValueError("Confirm that ", img, " exists")
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))
if len(faces) > 0:
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)
# ---------------------------
# face alignment
eyes = eye_detector.detectMultiScale(detected_face_gray)
if len(eyes) >= 2:
# find the largest 2 eye
base_eyes = eyes[:, 2]
items = []
for i in range(0, len(base_eyes)):
item = (base_eyes[i], i)
items.append(item)
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]
if eye_1[0] < eye_2[0]:
left_eye = eye_1
right_eye = eye_2
else:
left_eye = eye_2
right_eye = eye_1
# -----------------------
# find center of eyes
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]
right_eye_center = (
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]
# -----------------------
# find rotation direction
if left_eye_y > right_eye_y:
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
# -----------------------
# find length of triangle edges
a = distance(left_eye_center, point_3rd)
b = distance(right_eye_center, point_3rd)
c = distance(right_eye_center, left_eye_center)
# -----------------------
# apply cosine rule
cos_a = (b * b + c * c - a * a) / (2 * b * c)
angle = np.arccos(cos_a) # angle in radian
angle = (angle * 180) / math.pi # radian to degree
# -----------------------
# rotate base image
if direction == -1:
angle = 90 - angle
img = Image.fromarray(img_raw)
img = np.array(img.rotate(direction * angle))
# 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.
if grayscale == True:
detected_face = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
detected_face = cv2.resize(detected_face, target_size)
img_pixels = image.img_to_array(detected_face)
img_pixels = np.expand_dims(img_pixels, axis=0)
# normalize input in [0, 1]
img_pixels /= 255
return img_pixels
else:
if (exact_image == True) or (enforce_detection != True):
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."
)
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
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:
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:
img = loadBase64Img(img)
elif exact_image != True: #image path passed as input
if os.path.isfile(img) != True:
raise ValueError("Confirm that ",img," exists")
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))
if len(faces) > 0:
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)
#---------------------------
#face alignment
eyes = eye_detector.detectMultiScale(detected_face_gray)
if len(eyes) >= 2:
#find the largest 2 eye
base_eyes = eyes[:, 2]
items = []
for i in range(0, len(base_eyes)):
item = (base_eyes[i], i)
items.append(item)
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]
if eye_1[0] < eye_2[0]:
left_eye = eye_1
right_eye = eye_2
else:
left_eye = eye_2
right_eye = eye_1
#-----------------------
#find center of eyes
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]
right_eye_center = (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]
#-----------------------
#find rotation direction
if left_eye_y > right_eye_y:
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
#-----------------------
#find length of triangle edges
a = distance(left_eye_center, point_3rd)
b = distance(right_eye_center, point_3rd)
c = distance(right_eye_center, left_eye_center)
#-----------------------
#apply cosine rule
cos_a = (b*b + c*c - a*a)/(2*b*c)
angle = np.arccos(cos_a) #angle in radian
angle = (angle * 180) / math.pi #radian to degree
#-----------------------
#rotate base image
if direction == -1:
angle = 90 - angle
img = Image.fromarray(img_raw)
img = np.array(img.rotate(direction * angle))
#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.
if grayscale == True:
detected_face = cv2.cvtColor(detected_face, cv2.COLOR_BGR2GRAY)
detected_face = cv2.resize(detected_face, target_size)
img_pixels = image.img_to_array(detected_face)
img_pixels = np.expand_dims(img_pixels, axis = 0)
#normalize input in [0, 1]
img_pixels /= 255
return img_pixels
else:
if (exact_image == True) or (enforce_detection != True):
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.")
def allocateMemory():
#find allocated memories
gpu_indexes = []
memory_usage_percentages = []; available_memories = []; total_memories = []; utilizations = []
power_usages = []; power_capacities = []
try:
result = subprocess.check_output(['nvidia-smi'])
dashboard = result.decode("utf-8").split("=|")
# find allocated memories
gpu_indexes = []
memory_usage_percentages = []
available_memories = []
total_memories = []
utilizations = []
power_usages = []
power_capacities = []
dashboard = dashboard[1].split("\n")
gpu_idx = 0
for line in dashboard:
if ("MiB" in line):
power_info = line.split("|")[1]
power_capacity = int(power_info.split("/")[-1].replace("W", ""))
power_usage = int((power_info.split("/")[-2]).strip().split(" ")[-1].replace("W", ""))
power_usages.append(power_usage)
power_capacities.append(power_capacity)
#----------------------------
memory_info = line.split("|")[2].replace("MiB","").split("/")
utilization_info = int(line.split("|")[3].split("%")[0])
allocated = int(memory_info[0])
total_memory = int(memory_info[1])
available_memory = total_memory - allocated
total_memories.append(total_memory)
available_memories.append(available_memory)
memory_usage_percentages.append(round(100*int(allocated)/int(total_memory), 4))
utilizations.append(utilization_info)
gpu_indexes.append(gpu_idx)
gpu_idx = gpu_idx + 1
gpu_count = gpu_idx * 1
except Exception as err:
gpu_count = 0
#print(str(err))
#------------------------------
df = pd.DataFrame(gpu_indexes, columns = ["gpu_index"])
df["total_memories_in_mb"] = total_memories
df["available_memories_in_mb"] = available_memories
df["memory_usage_percentage"] = memory_usage_percentages
df["utilizations"] = utilizations
df["power_usages_in_watts"] = power_usages
df["power_capacities_in_watts"] = power_capacities
df = df.sort_values(by = ["available_memories_in_mb"], ascending = False).reset_index(drop = True)
#------------------------------
required_memory = 10000 #All deepface models require 9016 MiB
if df.shape[0] > 0: #has gpu
if df.iloc[0].available_memories_in_mb > required_memory:
my_gpu = str(int(df.iloc[0].gpu_index))
os.environ["CUDA_VISIBLE_DEVICES"] = my_gpu
#------------------------------
#tf allocates all memory by default
#this block avoids greedy approach
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
keras.backend.set_session(session)
print("DeepFace will run on GPU (gpu_", my_gpu,")")
else:
#this case has gpu but no enough memory to allocate
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("DeepFace will run on CPU")
else:
print("DeepFace will run on CPU")
#------------------------------
try:
result = subprocess.check_output(["nvidia-smi"])
dashboard = result.decode("utf-8").split("=|")
dashboard = dashboard[1].split("\n")
gpu_idx = 0
for line in dashboard:
if "MiB" in line:
power_info = line.split("|")[1]
power_capacity = int(power_info.split("/")[-1].replace("W", ""))
power_usage = int(
(power_info.split("/")[-2]).strip().split(" ")[-1].replace("W", "")
)
power_usages.append(power_usage)
power_capacities.append(power_capacity)
# ----------------------------
memory_info = line.split("|")[2].replace("MiB", "").split("/")
utilization_info = int(line.split("|")[3].split("%")[0])
allocated = int(memory_info[0])
total_memory = int(memory_info[1])
available_memory = total_memory - allocated
total_memories.append(total_memory)
available_memories.append(available_memory)
memory_usage_percentages.append(
round(100 * int(allocated) / int(total_memory), 4)
)
utilizations.append(utilization_info)
gpu_indexes.append(gpu_idx)
gpu_idx = gpu_idx + 1
gpu_count = gpu_idx * 1
except Exception as err:
gpu_count = 0
# print(str(err))
# ------------------------------
df = pd.DataFrame(gpu_indexes, columns=["gpu_index"])
df["total_memories_in_mb"] = total_memories
df["available_memories_in_mb"] = available_memories
df["memory_usage_percentage"] = memory_usage_percentages
df["utilizations"] = utilizations
df["power_usages_in_watts"] = power_usages
df["power_capacities_in_watts"] = power_capacities
df = df.sort_values(by=["available_memories_in_mb"], ascending=False).reset_index(
drop=True
)
# ------------------------------
required_memory = 10000 # All deepface models require 9016 MiB
if df.shape[0] > 0: # has gpu
if df.iloc[0].available_memories_in_mb > required_memory:
my_gpu = str(int(df.iloc[0].gpu_index))
os.environ["CUDA_VISIBLE_DEVICES"] = my_gpu
# ------------------------------
# tf allocates all memory by default
# this block avoids greedy approach
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
keras.backend.set_session(session)
print("DeepFace will run on GPU (gpu_", my_gpu, ")")
else:
# this case has gpu but no enough memory to allocate
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("DeepFace will run on CPU")
else:
print("DeepFace will run on CPU")
# ------------------------------