Merge pull request #135 from ma7555/master

Loads MTCNN once only + convert to RGB before inferenence + add better readibility

deepface/DeepFace.py

@@ -10,18 +10,20 @@ import numpy as np
 import pandas as pd
 from tqdm import tqdm
 import json
-import cv2
-from keras import backend as K
-import keras
-import tensorflow as tf
 import pickle
 
-from deepface import DeepFace
 from deepface.basemodels import VGGFace, OpenFace, Facenet, FbDeepFace, DeepID
 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, detector_backend = 'opencv'):
+def DlibResNet_():
+	from deepface.basemodels.DlibResNet import DlibResNet
+	return DlibResNet()
+
+def verify(img1_path, img2_path = '', model_name='VGG-Face', distance_metric='cosine', 
+		   model=None, enforce_detection=True, detector_backend = 'opencv'):
+
+	# this is not a must because it is very huge.		
 
 	tic = time.time()
 
@@ -34,6 +36,9 @@ def verify(img1_path, img2_path = '', model_name ='VGG-Face', distance_metric =
 
 	#------------------------------
 	
+	if detector_backend == 'mtcnn':
+		functions.load_mtcnn()
+
 	resp_objects = []
 	
 	if model_name == 'Ensemble':
@@ -199,33 +204,23 @@ def verify(img1_path, img2_path = '', model_name ='VGG-Face', distance_metric =
 	#ensemble learning disabled
 	
 	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':
+		models = {
-			print("Using OpenFace model backend", distance_metric,"distance.")
+				'VGG-Face': VGGFace.loadModel, 
-			model = OpenFace.loadModel()
+				'OpenFace': OpenFace.loadModel,
-
+				'Facenet': Facenet.loadModel,
-		elif model_name == 'Facenet':
+				'DeepFace': FbDeepFace.loadModel,
-			print("Using Facenet model backend", distance_metric,"distance.")
+				'DeepID': DeepID.loadModel,
-			model = Facenet.loadModel()
+				'Dlib': DlibResNet_
-
+				}
-		elif model_name == 'DeepFace':
-			print("Using FB DeepFace model backend", distance_metric,"distance.")
-			model = FbDeepFace.loadModel()
-		
-		elif model_name == 'DeepID':
-			print("Using DeepID2 model backend", distance_metric,"distance.")
-			model = DeepID.loadModel()
-		
-		elif model_name == 'Dlib':
-			print("Using Dlib ResNet model backend", distance_metric,"distance.")
-			from deepface.basemodels.DlibResNet import DlibResNet #this is not a must because it is very huge.
-			model = DlibResNet()
 
+		model = models.get(model_name)
+		if model:
+			model = model()
+			print('Using {} model backend and {} distance'.format(model_name, distance_metric))
 		else:
-			raise ValueError("Invalid model_name passed - ", model_name)
+			raise ValueError('Invalid model_name passed - {}'.format(model_name))
+	
 	else: #model != None
 		print("Already built model is passed")
 
@@ -344,7 +339,6 @@ def verify(img1_path, img2_path = '', model_name ='VGG-Face', distance_metric =
 		return resp_obj
 		#return resp_objects
 
-
 def analyze(img_path, actions = [], models = {}, enforce_detection = True, detector_backend = 'opencv'):
 
 	if type(img_path) == list:
@@ -513,7 +507,6 @@ def analyze(img_path, actions = [], models = {}, enforce_detection = True, detec
 		return resp_obj
 		#return resp_objects
 
-
 def detectFace(img_path, detector_backend = 'opencv'):
 	img = functions.preprocess_face(img = img_path, detector_backend = detector_backend)[0] #preprocess_face returns (1, 224, 224, 3)
 	return img[:, :, ::-1] #bgr to rgb

deepface/commons/functions.py

@@ -20,6 +20,10 @@ import bz2
 from deepface.commons import distance
 from mtcnn import MTCNN #0.1.0
 
+def load_mtcnn():
+	global mtcnn_detector
+	mtcnn_detector = MTCNN()
+
 def loadBase64Img(uri):
    encoded_data = uri.split(',')[1]
    nparr = np.fromstring(base64.b64decode(encoded_data), np.uint8)
@@ -40,55 +44,18 @@ def initializeFolder():
 	
 def findThreshold(model_name, distance_metric):
 	
-	threshold = 0.40
+	base_threshold = {'cosine': 0.40, 'euclidean': 0.55, 'euclidean_l2': 0.75}
 	
-	if model_name == 'VGG-Face':
+	thresholds = {
-		if distance_metric == 'cosine':
+		'VGG-Face': {'cosine': 0.40, 'euclidean': 0.55, 'euclidean_l2': 0.75},
-			threshold = 0.40
+		'OpenFace': {'cosine': 0.10, 'euclidean': 0.55, 'euclidean_l2': 0.55},
-		elif distance_metric == 'euclidean':
+		'Facenet':  {'cosine': 0.40, 'euclidean': 10, 'euclidean_l2': 0.80},
-			threshold = 0.55
+		'DeepFace': {'cosine': 0.23, 'euclidean': 64, 'euclidean_l2': 0.64},
-		elif distance_metric == 'euclidean_l2':
+		'DeepID': 	{'cosine': 0.015, 'euclidean': 45, 'euclidean_l2': 0.17},
-			threshold = 0.75	
+		'Dlib': 	{'cosine': 0.07, 'euclidean': 0.6, 'euclidean_l2': 0.6}
+		}
 
-	elif model_name == 'OpenFace':
+	threshold = thresholds.get(model_name, base_threshold).get(distance_metric, 0.4)
-		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
-	
-	elif model_name == 'DeepID':
-		if distance_metric == 'cosine':
-			threshold = 0.015
-		elif distance_metric == 'euclidean':
-			threshold = 45
-		elif distance_metric == 'euclidean_l2':
-			threshold = 0.17
-	
-	elif model_name == 'Dlib':
-		if distance_metric == 'cosine':
-			threshold = 0.07
-		elif distance_metric == 'euclidean':
-			threshold = 0.60
-		elif distance_metric == 'euclidean_l2':
-			threshold = 0.60
 	
 	return threshold
 
@@ -277,9 +244,9 @@ def detect_face(img, detector_backend = 'opencv', grayscale = False, enforce_det
 		
 	elif detector_backend == 'mtcnn':
 		
-		mtcnn_detector = MTCNN()
+		# mtcnn_detector = MTCNN()
-		
+		img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
-		detections = mtcnn_detector.detect_faces(img)
+		detections = mtcnn_detector.detect_faces(img_rgb)
 		
 		if len(detections) > 0:
 			detection = detections[0]
@@ -288,7 +255,7 @@ def detect_face(img, detector_backend = 'opencv', grayscale = False, enforce_det
 			return detected_face
 		
 		else: #if no face detected
-			if enforce_detection != True:			
+			if not enforce_detection:			
 				return img
 	
 			else:
@@ -432,8 +399,9 @@ def align_face(img, detector_backend = 'opencv'):
 	
 	elif detector_backend == 'mtcnn':
 		
-		mtcnn_detector = MTCNN()
+		# mtcnn_detector = MTCNN()
-		detections = mtcnn_detector.detect_faces(img)
+		img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+		detections = mtcnn_detector.detect_faces(img_rgb)
 		
 		if len(detections) > 0:
 			detection = detections[0]