github/deepface
1
1
Fork 0
mirror of https://github.com/serengil/deepface.git synced 2025-06-07 12:05:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

clean code refactoring

Browse Source
This commit is contained in:
Sefik Ilkin Serengil 2023-01-29 21:39:54 +00:00
parent 029f3dfa43
commit 2d39ade138
5 changed files with 255 additions and 262 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
deepface/DeepFace.py
View File

@ -26,9 +26,6 @@ from deepface.basemodels import (
SFace, SFace,
) )
from deepface.extendedmodels import Age, Gender, Race, Emotion from deepface.extendedmodels import Age, Gender, Race, Emotion
from deepface.extendedmodels.Emotion import EMOTION_LABELS
from deepface.extendedmodels.Gender import GENDER_LABELS
from deepface.extendedmodels.Race import RACE_LABELS
from deepface.commons import functions, realtime, distance as dst from deepface.commons import functions, realtime, distance as dst
# ----------------------------------- # -----------------------------------
@ -342,11 +339,11 @@ def analyze(
obj["emotion"] = {} obj["emotion"] = {}
for i, emotion_label in enumerate(EMOTION_LABELS): for i, emotion_label in enumerate(Emotion.labels):
emotion_prediction = 100 * emotion_predictions[i] / sum_of_predictions emotion_prediction = 100 * emotion_predictions[i] / sum_of_predictions
obj["emotion"][emotion_label] = emotion_prediction obj["emotion"][emotion_label] = emotion_prediction
obj["dominant_emotion"] = EMOTION_LABELS[np.argmax(emotion_predictions)] obj["dominant_emotion"] = Emotion.labels[np.argmax(emotion_predictions)]
elif action == "age": elif action == "age":
age_predictions = models["age"].predict(img_content, verbose=0)[0, :] age_predictions = models["age"].predict(img_content, verbose=0)[0, :]
@ -357,22 +354,22 @@ def analyze(
elif action == "gender": elif action == "gender":
gender_predictions = models["gender"].predict(img_content, verbose=0)[0, :] gender_predictions = models["gender"].predict(img_content, verbose=0)[0, :]
obj["gender"] = {} obj["gender"] = {}
for i, gender_label in enumerate(GENDER_LABELS): for i, gender_label in enumerate(Gender.labels):
gender_prediction = 100 * gender_predictions[i] gender_prediction = 100 * gender_predictions[i]
obj["gender"][gender_label] = gender_prediction obj["gender"][gender_label] = gender_prediction
obj["dominant_gender"] = GENDER_LABELS[np.argmax(gender_predictions)] obj["dominant_gender"] = Gender.labels[np.argmax(gender_predictions)]
elif action == "race": elif action == "race":
race_predictions = models["race"].predict(img_content, verbose=0)[0, :] race_predictions = models["race"].predict(img_content, verbose=0)[0, :]
sum_of_predictions = race_predictions.sum() sum_of_predictions = race_predictions.sum()
obj["race"] = {} obj["race"] = {}
for i, race_label in enumerate(RACE_LABELS): for i, race_label in enumerate(Race.labels):
race_prediction = 100 * race_predictions[i] / sum_of_predictions race_prediction = 100 * race_predictions[i] / sum_of_predictions
obj["race"][race_label] = race_prediction obj["race"][race_label] = race_prediction
obj["dominant_race"] = RACE_LABELS[np.argmax(race_predictions)] obj["dominant_race"] = Race.labels[np.argmax(race_predictions)]
# ----------------------------- # -----------------------------
# mention facial areas # mention facial areas

476
deepface/commons/realtime.py
View File

@ -4,7 +4,7 @@ import numpy as np
import pandas as pd import pandas as pd
import cv2 import cv2
from deepface import DeepFace from deepface import DeepFace
from deepface.commons import functions, distance as dst from deepface.commons import functions
# dependency configuration # dependency configuration
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2" os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
@ -30,7 +30,6 @@ def analysis(
enable_age_gender = True enable_age_gender = True
# ------------------------ # ------------------------
# find custom values for this input set # find custom values for this input set
threshold = dst.findThreshold(model_name, distance_metric)
target_size = functions.find_target_size(model_name=model_name) target_size = functions.find_target_size(model_name=model_name)
# ------------------------ # ------------------------
# build models once to store them in the memory # build models once to store them in the memory
@ -423,260 +422,259 @@ def analysis(
if df.shape[0] > 0: if df.shape[0] > 0:
candidate = df.iloc[0] candidate = df.iloc[0]
label = candidate["identity"] label = candidate["identity"]
best_distance = candidate[f"{model_name}_{distance_metric}"]
if best_distance <= threshold:
# to use this source image as is
display_img = cv2.imread(label)
# to use extracted face
source_objs = DeepFace.extract_faces(
img_path=label,
target_size=(pivot_img_size, pivot_img_size),
detector_backend=detector_backend,
enforce_detection=False,
align=False,
)
if len(source_objs) > 0: # to use this source image as is
# extract 1st item directly display_img = cv2.imread(label)
source_obj = source_objs[0] # to use extracted face
display_img = source_obj["face"] source_objs = DeepFace.extract_faces(
display_img *= 255 img_path=label,
display_img = display_img[:, :, ::-1] target_size=(pivot_img_size, pivot_img_size),
# -------------------- detector_backend=detector_backend,
label = label.split("/")[-1] enforce_detection=False,
align=False,
)
try: if len(source_objs) > 0:
if ( # extract 1st item directly
y - pivot_img_size > 0 source_obj = source_objs[0]
and x + w + pivot_img_size < resolution_x display_img = source_obj["face"]
): display_img *= 255
# top right display_img = display_img[:, :, ::-1]
freeze_img[ # --------------------
y - pivot_img_size : y, label = label.split("/")[-1]
x + w : x + w + pivot_img_size,
] = display_img
overlay = freeze_img.copy() try:
opacity = 0.4 if (
cv2.rectangle( y - pivot_img_size > 0
freeze_img, and x + w + pivot_img_size < resolution_x
(x + w, y), ):
(x + w + pivot_img_size, y + 20), # top right
(46, 200, 255), freeze_img[
cv2.FILLED, y - pivot_img_size : y,
) x + w : x + w + pivot_img_size,
cv2.addWeighted( ] = display_img
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
cv2.putText( overlay = freeze_img.copy()
freeze_img, opacity = 0.4
label, cv2.rectangle(
(x + w, y + 10), freeze_img,
cv2.FONT_HERSHEY_SIMPLEX, (x + w, y),
0.5, (x + w + pivot_img_size, y + 20),
text_color, (46, 200, 255),
1, cv2.FILLED,
) )
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
# connect face and text cv2.putText(
cv2.line( freeze_img,
freeze_img, label,
(x + int(w / 2), y), (x + w, y + 10),
(x + 3 * int(w / 4), y - int(pivot_img_size / 2)), cv2.FONT_HERSHEY_SIMPLEX,
(67, 67, 67), 0.5,
1, text_color,
) 1,
cv2.line( )
freeze_img,
(x + 3 * int(w / 4), y - int(pivot_img_size / 2)),
(x + w, y - int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
elif ( # connect face and text
y + h + pivot_img_size < resolution_y cv2.line(
and x - pivot_img_size > 0 freeze_img,
): (x + int(w / 2), y),
# bottom left (x + 3 * int(w / 4), y - int(pivot_img_size / 2)),
freeze_img[ (67, 67, 67),
y + h : y + h + pivot_img_size, 1,
x - pivot_img_size : x, )
] = display_img cv2.line(
freeze_img,
(x + 3 * int(w / 4), y - int(pivot_img_size / 2)),
(x + w, y - int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
overlay = freeze_img.copy() elif (
opacity = 0.4 y + h + pivot_img_size < resolution_y
cv2.rectangle( and x - pivot_img_size > 0
freeze_img, ):
(x - pivot_img_size, y + h - 20), # bottom left
(x, y + h), freeze_img[
(46, 200, 255), y + h : y + h + pivot_img_size,
cv2.FILLED, x - pivot_img_size : x,
) ] = display_img
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
cv2.putText( overlay = freeze_img.copy()
freeze_img, opacity = 0.4
label, cv2.rectangle(
(x - pivot_img_size, y + h - 10), freeze_img,
cv2.FONT_HERSHEY_SIMPLEX, (x - pivot_img_size, y + h - 20),
0.5, (x, y + h),
text_color, (46, 200, 255),
1, cv2.FILLED,
) )
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
# connect face and text cv2.putText(
cv2.line( freeze_img,
freeze_img, label,
(x + int(w / 2), y + h), (x - pivot_img_size, y + h - 10),
( cv2.FONT_HERSHEY_SIMPLEX,
x + int(w / 2) - int(w / 4), 0.5,
y + h + int(pivot_img_size / 2), text_color,
), 1,
(67, 67, 67), )
1,
)
cv2.line(
freeze_img,
(
x + int(w / 2) - int(w / 4),
y + h + int(pivot_img_size / 2),
),
(x, y + h + int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
elif y - pivot_img_size > 0 and x - pivot_img_size > 0: # connect face and text
# top left cv2.line(
freeze_img[ freeze_img,
y - pivot_img_size : y, x - pivot_img_size : x (x + int(w / 2), y + h),
] = display_img (
x + int(w / 2) - int(w / 4),
y + h + int(pivot_img_size / 2),
),
(67, 67, 67),
1,
)
cv2.line(
freeze_img,
(
x + int(w / 2) - int(w / 4),
y + h + int(pivot_img_size / 2),
),
(x, y + h + int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
overlay = freeze_img.copy() elif y - pivot_img_size > 0 and x - pivot_img_size > 0:
opacity = 0.4 # top left
cv2.rectangle( freeze_img[
freeze_img, y - pivot_img_size : y, x - pivot_img_size : x
(x - pivot_img_size, y), ] = display_img
(x, y + 20),
(46, 200, 255),
cv2.FILLED,
)
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
cv2.putText( overlay = freeze_img.copy()
freeze_img, opacity = 0.4
label, cv2.rectangle(
(x - pivot_img_size, y + 10), freeze_img,
cv2.FONT_HERSHEY_SIMPLEX, (x - pivot_img_size, y),
0.5, (x, y + 20),
text_color, (46, 200, 255),
1, cv2.FILLED,
) )
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
# connect face and text cv2.putText(
cv2.line( freeze_img,
freeze_img, label,
(x + int(w / 2), y), (x - pivot_img_size, y + 10),
( cv2.FONT_HERSHEY_SIMPLEX,
x + int(w / 2) - int(w / 4), 0.5,
y - int(pivot_img_size / 2), text_color,
), 1,
(67, 67, 67), )
1,
)
cv2.line(
freeze_img,
(
x + int(w / 2) - int(w / 4),
y - int(pivot_img_size / 2),
),
(x, y - int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
elif ( # connect face and text
x + w + pivot_img_size < resolution_x cv2.line(
and y + h + pivot_img_size < resolution_y freeze_img,
): (x + int(w / 2), y),
# bottom righ (
freeze_img[ x + int(w / 2) - int(w / 4),
y + h : y + h + pivot_img_size, y - int(pivot_img_size / 2),
x + w : x + w + pivot_img_size, ),
] = display_img (67, 67, 67),
1,
)
cv2.line(
freeze_img,
(
x + int(w / 2) - int(w / 4),
y - int(pivot_img_size / 2),
),
(x, y - int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
overlay = freeze_img.copy() elif (
opacity = 0.4 x + w + pivot_img_size < resolution_x
cv2.rectangle( and y + h + pivot_img_size < resolution_y
freeze_img, ):
(x + w, y + h - 20), # bottom righ
(x + w + pivot_img_size, y + h), freeze_img[
(46, 200, 255), y + h : y + h + pivot_img_size,
cv2.FILLED, x + w : x + w + pivot_img_size,
) ] = display_img
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
cv2.putText( overlay = freeze_img.copy()
freeze_img, opacity = 0.4
label, cv2.rectangle(
(x + w, y + h - 10), freeze_img,
cv2.FONT_HERSHEY_SIMPLEX, (x + w, y + h - 20),
0.5, (x + w + pivot_img_size, y + h),
text_color, (46, 200, 255),
1, cv2.FILLED,
) )
cv2.addWeighted(
overlay,
opacity,
freeze_img,
1 - opacity,
0,
freeze_img,
)
# connect face and text cv2.putText(
cv2.line( freeze_img,
freeze_img, label,
(x + int(w / 2), y + h), (x + w, y + h - 10),
( cv2.FONT_HERSHEY_SIMPLEX,
x + int(w / 2) + int(w / 4), 0.5,
y + h + int(pivot_img_size / 2), text_color,
), 1,
(67, 67, 67), )
1,
) # connect face and text
cv2.line( cv2.line(
freeze_img, freeze_img,
( (x + int(w / 2), y + h),
x + int(w / 2) + int(w / 4), (
y + h + int(pivot_img_size / 2), x + int(w / 2) + int(w / 4),
), y + h + int(pivot_img_size / 2),
(x + w, y + h + int(pivot_img_size / 2)), ),
(67, 67, 67), (67, 67, 67),
1, 1,
) )
except Exception as err: # pylint: disable=broad-except cv2.line(
print(str(err)) freeze_img,
(
x + int(w / 2) + int(w / 4),
y + h + int(pivot_img_size / 2),
),
(x + w, y + h + int(pivot_img_size / 2)),
(67, 67, 67),
1,
)
except Exception as err: # pylint: disable=broad-except
print(str(err))
tic = time.time() # in this way, freezed image can show 5 seconds tic = time.time() # in this way, freezed image can show 5 seconds

7
deepface/extendedmodels/Emotion.py
View File

@ -24,6 +24,9 @@ elif tf_version == 2:
) )
# ------------------------------------------- # -------------------------------------------
# Labels for the emotions that can be detected by the model.
labels = ["angry", "disgust", "fear", "happy", "sad", "surprise", "neutral"]
def loadModel( def loadModel(
url="https://github.com/serengil/deepface_models/releases/download/v1.0/facial_expression_model_weights.h5", url="https://github.com/serengil/deepface_models/releases/download/v1.0/facial_expression_model_weights.h5",
@ -70,7 +73,3 @@ def loadModel(
model.load_weights(home + "/.deepface/weights/facial_expression_model_weights.h5") model.load_weights(home + "/.deepface/weights/facial_expression_model_weights.h5")
return model return model
EMOTION_LABELS = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral']
"""Labels for the emotions that can be detected by the model."""

11
deepface/extendedmodels/Gender.py
View File

@ -18,6 +18,11 @@ elif tf_version == 2:
from tensorflow.keras.models import Model, Sequential from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.layers import Convolution2D, Flatten, Activation from tensorflow.keras.layers import Convolution2D, Flatten, Activation
# ------------------------------------- # -------------------------------------
# Labels for the genders that can be detected by the model.
labels = ["Woman", "Man"]
def loadModel( def loadModel(
url="https://github.com/serengil/deepface_models/releases/download/v1.0/gender_model_weights.h5", url="https://github.com/serengil/deepface_models/releases/download/v1.0/gender_model_weights.h5",
): ):
@ -51,9 +56,3 @@ def loadModel(
gender_model.load_weights(home + "/.deepface/weights/gender_model_weights.h5") gender_model.load_weights(home + "/.deepface/weights/gender_model_weights.h5")
return gender_model return gender_model
# --------------------------
GENDER_LABELS = ["Woman", "Man"]
"""Labels for the genders that can be detected by the model."""

8
deepface/extendedmodels/Race.py
View File

@ -17,6 +17,10 @@ elif tf_version == 2:
from tensorflow.keras.models import Model, Sequential from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.layers import Convolution2D, Flatten, Activation from tensorflow.keras.layers import Convolution2D, Flatten, Activation
# -------------------------- # --------------------------
# Labels for the ethnic phenotypes that can be detected by the model.
labels = ["asian", "indian", "black", "white", "middle eastern", "latino hispanic"]
def loadModel( def loadModel(
url="https://github.com/serengil/deepface_models/releases/download/v1.0/race_model_single_batch.h5", url="https://github.com/serengil/deepface_models/releases/download/v1.0/race_model_single_batch.h5",
): ):
@ -50,7 +54,3 @@ def loadModel(
race_model.load_weights(home + "/.deepface/weights/race_model_single_batch.h5") race_model.load_weights(home + "/.deepface/weights/race_model_single_batch.h5")
return race_model return race_model
RACE_LABELS = ['asian', 'indian', 'black', 'white', 'middle eastern', 'latino hispanic']
"""Labels for the ethnic phenotypes that can be detected by the model."""