model building once for api

api/app.py

@@ -1,121 +0,0 @@
-from flask import Flask, jsonify, request, make_response
-
-import uuid
-import json
-import time
-
-from deepface import DeepFace
-
-#------------------------------
-
-app = Flask(__name__)
-
-#------------------------------
-#Service API Interface
-
-@app.route('/')
-def index():
-	return '<h1>Hello, world!</h1>'
-
-@app.route('/analyze', methods=['POST'])
-def analyze():
-
-	req = request.get_json()
-	trx_id = uuid.uuid4()
-
-	#---------------------------
-
-	tic = time.time()
-
-	instances = []
-	if "img" in list(req.keys()):
-		raw_content = req["img"] #list
-
-		for item in raw_content: #item is in type of dict
-			instances.append(item)
-	
-	if len(instances) == 0:
-		return jsonify({'success': False, 'error': 'you must pass at least one img object in your request'}), 205
-	
-	print("Analyzing ", len(instances)," instances")
-
-	#---------------------------
-
-	actions= ['emotion', 'age', 'gender', 'race']
-	if "actions" in list(req.keys()):
-		actions = req["actions"]
-	
-	#---------------------------
-
-	resp_obj = DeepFace.analyze(instances, actions=actions)
-	
-	#---------------------------
-
-	toc = time.time()
-
-	resp_obj["trx_id"] = trx_id
-	resp_obj["seconds"] = toc-tic
-
-	return resp_obj
-
-@app.route('/verify', methods=['POST'])
-
-def verify():
-
-	req = request.get_json()
-	trx_id = uuid.uuid4()
-
-	tic = time.time()
-	
-	#-------------------------
-	
-	model_name = "VGG-Face"; distance_metric = "cosine"
-	if "model_name" in list(req.keys()):
-		model_name = req["model_name"]
-	if "distance_metric" in list(req.keys()):
-		distance_metric = req["distance_metric"]
-	
-	instances = []
-	if "img" in list(req.keys()):
-		raw_content = req["img"] #list
-
-		for item in raw_content: #item is in type of dict
-			instance = []
-			img1 = item["img1"]; img2 = item["img2"]
-
-			validate_img1 = False
-			if len(img1) > 11 and img1[0:11] == "data:image/":
-				validate_img1 = True
-			
-			validate_img2 = False
-			if len(img2) > 11 and img2[0:11] == "data:image/":
-				validate_img2 = True
-
-			if validate_img1 != True or validate_img2 != True:
-				return jsonify({'success': False, 'error': 'you must pass both img1 and img2 as base64 encoded string'}), 205
-
-			instance.append(img1); instance.append(img2)
-			instances.append(instance)
-		
-	#--------------------------
-
-	if len(instances) == 0:
-		return jsonify({'success': False, 'error': 'you must pass at least one img object in your request'}), 205
-	
-	print("Input request of ", trx_id, " has ",len(instances)," pairs to verify")
-	
-	#--------------------------
-	resp_obj = DeepFace.verify(instances, model_name = model_name, distance_metric = distance_metric)
-
-	toc =  time.time()
-
-	resp_obj["trx_id"] = trx_id
-	resp_obj["seconds"] = toc-tic
-
-	#--------------------------
-	
-	return resp_obj, 200
-
-if __name__ == '__main__':
-	
-	app.run()

deepface/DeepFace.py

@@ -9,6 +9,8 @@ from tqdm import tqdm
 import json
 import cv2
 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
@@ -18,15 +20,11 @@ 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
 
-#TO-DO: pass built model optionally as input. I will get complex models up in rest api once and call these functions directly
-
 def verify(img1_path, img2_path=''
-	, model_name ='VGG-Face', distance_metric = 'cosine', plot = False):
+	, model_name ='VGG-Face', distance_metric = 'cosine', model = None):
 	
 	tic = time.time()
 	
-	K.clear_session()
-	
 	if type(img1_path) == list:
 		bulkProcess = True
 		img_list = img1_path.copy()
@@ -36,28 +34,31 @@ def verify(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()
-		input_shape = (224, 224)	
 		
 		elif model_name == 'OpenFace':
 			print("Using OpenFace model backend", distance_metric,"distance.")
 			model = OpenFace.loadModel()
-		input_shape = (96, 96)
 		
 		elif model_name == 'Facenet':
 			print("Using Facenet model backend", distance_metric,"distance.")
 			model = Facenet.loadModel()
-		input_shape = (160, 160)
 		
 		elif model_name == 'DeepFace':
 			print("Using FB DeepFace model backend", distance_metric,"distance.")
 			model = FbDeepFace.loadModel()
-		input_shape = (152, 152)
 		
 		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]
 
 	#------------------------------
 	
@@ -133,8 +134,6 @@ def verify(img1_path, img2_path=''
 	#print("identification lasts ",toc-tic," seconds")
 	
 	if bulkProcess == True:
-		K.clear_session()
-
 		resp_obj = "{"
 		
 		for i in range(0, len(resp_objects)):
@@ -151,8 +150,6 @@ def verify(img1_path, img2_path=''
 
 def analyze(img_path, actions= []):
 
-	K.clear_session()
-	
 	if type(img_path) == list:
 		img_paths = img_path.copy()
 		bulkProcess = True
@@ -275,12 +272,9 @@ def analyze(img_path, actions= []):
 		if bulkProcess == True:
 			resp_objects.append(resp_obj)
 		else:
-			K.clear_session()
 			return resp_obj
 	
 	if bulkProcess == True:
-		K.clear_session()
-
 		resp_obj = "{"
 		
 		for i in range(0, len(resp_objects)):

deepface/app.py

@@ -0,0 +1,165 @@
+from flask import Flask, jsonify, request, make_response
+
+import uuid
+import json
+import time
+
+import tensorflow as tf
+
+from deepface import DeepFace
+from deepface.basemodels import VGGFace, OpenFace, Facenet, FbDeepFace
+
+#import DeepFace
+#from basemodels import VGGFace, OpenFace, Facenet, FbDeepFace
+
+#------------------------------
+
+app = Flask(__name__)
+
+tic = time.time()
+
+vggface_model = VGGFace.loadModel()
+print("VGG-Face model is built.")
+
+openface_model = OpenFace.loadModel()
+print("OpenFace model is built")
+
+facenet_model = Facenet.loadModel()
+print("FaceNet model is built")
+
+deepface_model = FbDeepFace.loadModel()
+print("DeepFace model is built")
+
+toc = time.time()
+
+print("Face recognition models are built in ", toc-tic," seconds")
+
+graph = tf.get_default_graph()
+
+#------------------------------
+#Service API Interface
+
+@app.route('/')
+def index():
+	return '<h1>Hello, world!</h1>'
+
+@app.route('/analyze', methods=['POST'])
+def analyze():
+	
+	global graph
+	
+	tic = time.time()
+	req = request.get_json()
+	trx_id = uuid.uuid4()
+
+	#---------------------------
+	
+	resp_obj = jsonify({'success': False})
+	with graph.as_default():
+		instances = []
+		if "img" in list(req.keys()):
+			raw_content = req["img"] #list
+
+			for item in raw_content: #item is in type of dict
+				instances.append(item)
+		
+		if len(instances) == 0:
+			return jsonify({'success': False, 'error': 'you must pass at least one img object in your request'}), 205
+		
+		print("Analyzing ", len(instances)," instances")
+
+		#---------------------------
+
+		actions= ['emotion', 'age', 'gender', 'race']
+		if "actions" in list(req.keys()):
+			actions = req["actions"]
+		
+		#---------------------------
+
+		resp_obj = DeepFace.analyze(instances, actions=actions)
+		
+		#---------------------------
+
+	toc = time.time()
+
+	resp_obj["trx_id"] = trx_id
+	resp_obj["seconds"] = toc-tic
+
+	return resp_obj
+
+@app.route('/verify', methods=['POST'])
+
+def verify():
+	
+	global graph
+	
+	tic = time.time()
+	req = request.get_json()
+	trx_id = uuid.uuid4()
+	
+	resp_obj = jsonify({'success': False})
+	
+	with graph.as_default():
+		
+		model_name = "VGG-Face"; distance_metric = "cosine"
+		if "model_name" in list(req.keys()):
+			model_name = req["model_name"]
+		if "distance_metric" in list(req.keys()):
+			distance_metric = req["distance_metric"]
+		
+		#----------------------
+		
+		instances = []
+		if "img" in list(req.keys()):
+			raw_content = req["img"] #list
+
+			for item in raw_content: #item is in type of dict
+				instance = []
+				img1 = item["img1"]; img2 = item["img2"]
+
+				validate_img1 = False
+				if len(img1) > 11 and img1[0:11] == "data:image/":
+					validate_img1 = True
+				
+				validate_img2 = False
+				if len(img2) > 11 and img2[0:11] == "data:image/":
+					validate_img2 = True
+
+				if validate_img1 != True or validate_img2 != True:
+					return jsonify({'success': False, 'error': 'you must pass both img1 and img2 as base64 encoded string'}), 205
+
+				instance.append(img1); instance.append(img2)
+				instances.append(instance)
+			
+		#--------------------------
+
+		if len(instances) == 0:
+			return jsonify({'success': False, 'error': 'you must pass at least one img object in your request'}), 205
+		
+		print("Input request of ", trx_id, " has ",len(instances)," pairs to verify")
+		
+		#--------------------------
+		
+		if model_name == "VGG-Face":
+			resp_obj = DeepFace.verify(instances, model_name = model_name, distance_metric = distance_metric, model = vggface_model)
+		elif model_name == "Facenet":
+			resp_obj = DeepFace.verify(instances, model_name = model_name, distance_metric = distance_metric, model = facenet_model)
+		elif model_name == "OpenFace":
+			resp_obj = DeepFace.verify(instances, model_name = model_name, distance_metric = distance_metric, model = openface_model)
+		elif model_name == "DeepFace":
+			resp_obj = DeepFace.verify(instances, model_name = model_name, distance_metric = distance_metric, model = deepface_model)
+		else:
+			return jsonify({'success': False, 'error': 'You must pass a valid model name. Available models are VGG-Face, Facenet, OpenFace, DeepFace but you passed %s' % (model_name)}), 205
+		
+	#--------------------------
+	
+	toc =  time.time()
+	
+	resp_obj["trx_id"] = trx_id
+	resp_obj["seconds"] = toc-tic
+	
+	return resp_obj, 200
+
+if __name__ == '__main__':
+	
+	app.run()

deepface/basemodels/OpenFace.py

@@ -32,7 +32,7 @@ def loadModel():
 	x = Conv2D(192, (3, 3), name='conv3')(x)
 	x = BatchNormalization(axis=3, epsilon=0.00001, name='bn3')(x)
 	x = Activation('relu')(x)
-	Lambda(lambda x: tf.nn.lrn(x, alpha=1e-4, beta=0.75), name='lrn_2')(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)