changed model

src/align/detect_face.py

@@ -33,6 +33,8 @@
 #from math import floor
 import cv2
 import os
+import numba
+import ctypes
 
 def layer(op):
     """Decorator for composable network layers."""
@@ -289,12 +291,18 @@ def create_mtcnn(sess, model_path):
         data = tf.placeholder(tf.float32, (None,48,48,3), 'input')
         onet = ONet({'data':data})
         onet.load(os.path.join(model_path, 'det3.npy'), sess)
-
-    pnet_fun = lambda img : sess.run(('pnet/conv4-2/BiasAdd:0', 'pnet/prob1:0'), feed_dict={'pnet/input:0':img})
-    rnet_fun = lambda img : sess.run(('rnet/conv5-2/conv5-2:0', 'rnet/prob1:0'), feed_dict={'rnet/input:0':img})
-    onet_fun = lambda img : sess.run(('onet/conv6-2/conv6-2:0', 'onet/conv6-3/conv6-3:0', 'onet/prob1:0'), feed_dict={'onet/input:0':img})
+    sess.close()
+    with tf.device('/cpu:2'):
+        with tf.Graph().as_default():
+            tfconfig = tf.ConfigProto(device_count = {"CPU": 4}, allow_soft_placement = True)
+            tfconfig.intra_op_parallelism_threads = 8
+            tfconfig.inter_op_parallelism_threads = 8
+            with tf.Session(config = tfconfig) as sess:    
+                pnet_fun = lambda img : sess.run(('pnet/conv4-2/BiasAdd:0', 'pnet/prob1:0'), feed_dict={'pnet/input:0':img})
+                rnet_fun = lambda img : sess.run(('rnet/conv5-2/conv5-2:0', 'rnet/prob1:0'), feed_dict={'rnet/input:0':img})
+                onet_fun = lambda img : sess.run(('onet/conv6-2/conv6-2:0', 'onet/conv6-3/conv6-3:0', 'onet/prob1:0'), feed_dict={'onet/input:0':img})
     return pnet_fun, rnet_fun, onet_fun
-
+@numba.jit(nopython=True, parallel=True,fastmath = True)
 def detect_face(img, minsize, pnet, rnet, onet, threshold, factor):
     """Detects faces in an image, and returns bounding boxes and points for them.
     img: input image
@@ -643,6 +651,7 @@ def bulk_detect_face(images, detection_window_size_ratio, pnet, rnet, onet, thre
 
 
 # function [boundingbox] = bbreg(boundingbox,reg)
+@numba.jit(nopaython=True)
 def bbreg(boundingbox,reg):
     """Calibrate bounding boxes"""
     if reg.shape[1]==1:
@@ -656,7 +665,7 @@ def bbreg(boundingbox,reg):
     b4 = boundingbox[:,3]+reg[:,3]*h
     boundingbox[:,0:4] = np.transpose(np.vstack([b1, b2, b3, b4 ]))
     return boundingbox
- 
+@numba.jit(nopython=True, parallel=True,fastmath = True)
 def generateBoundingBox(imap, reg, scale, t):
     """Use heatmap to generate bounding boxes"""
     stride=2
@@ -682,8 +691,9 @@ def generateBoundingBox(imap, reg, scale, t):
     q2 = np.fix((stride*bb+cellsize-1+1)/scale)
     boundingbox = np.hstack([q1, q2, np.expand_dims(score,1), reg])
     return boundingbox, reg
- 
+
 # function pick = nms(boxes,threshold,type)
+@numba.jit(nopaython=True)
 def nms(boxes, threshold, method):
     if boxes.size==0:
         return np.empty((0,3))
@@ -717,6 +727,7 @@ def nms(boxes, threshold, method):
     return pick
 
 # function [dy edy dx edx y ey x ex tmpw tmph] = pad(total_boxes,w,h)
+@numba.jit(nopaython=True)
 def pad(total_boxes, w, h):
     """Compute the padding coordinates (pad the bounding boxes to square)"""
     tmpw = (total_boxes[:,2]-total_boxes[:,0]+1).astype(np.int32)
@@ -752,6 +763,7 @@ def pad(total_boxes, w, h):
     return dy, edy, dx, edx, y, ey, x, ex, tmpw, tmph
 
 # function [bboxA] = rerec(bboxA)
+@numba.jit(nopaython=True)
 def rerec(bboxA):
     """Convert bboxA to square."""
     h = bboxA[:,3]-bboxA[:,1]
@@ -761,7 +773,8 @@ def rerec(bboxA):
     bboxA[:,1] = bboxA[:,1]+h*0.5-l*0.5
     bboxA[:,2:4] = bboxA[:,0:2] + np.transpose(np.tile(l,(2,1)))
     return bboxA
-
+
+@numba.jit(nopaython=True)
 def imresample(img, sz):
     im_data = cv2.resize(img, (sz[1], sz[0]), interpolation=cv2.INTER_AREA) #@UndefinedVariable
     return im_data

src/data_processing.py

@@ -1,5 +1,5 @@
-"""
-process data and produce valid output for other moduls
+ """
+ process data and produce valid output for other moduls
 """
 import copy
 
@@ -71,7 +71,7 @@ def detect_faces_bbox(self, image):
         minsize = 30
         threshold = [self.pnet_threshold, self.rnet_threshold, self.onet_threshold]
         factor = 0.709
-        bounding_boxes, _ = align.detect_face.detect_face(image, minsize, self.pnet, self.rnet, self.onet, threshold,
+        boundin g_boxes, _ = align.detect_face.detect_face(image, minsize, self.pnet, self.rnet, self.onet, threshold,
                                                           factor)
         return bounding_boxes
 

src/deployment.py

@@ -4,13 +4,11 @@
 Handles ros operations necessary for using the package.
 """
 import os
-
 import roslib
 import rospkg
 import rospy
 from person_recognition.srv import *
 from std_msgs.msg import String
-
 from execution import Execution
 
 os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
@@ -69,7 +67,7 @@ def idle(self):
             rospy.spin()
         except KeyboardInterrupt:
             print("Shutting down person recognition module")
-            
+
     def dict_to_json_str(self, data):
         json_data = dict()
         for key, value in data.iteritems():

src/execution.py

@@ -12,7 +12,6 @@
 from data_acquisition import Warehouse
 from data_processing import DataProcessing
 from model_engineering import ModelEngineering
-
 import roslib
 import rospy
 import cv2

src/gender_detection.py

@@ -1,52 +0,0 @@
-from keras.preprocessing.image import img_to_array
-from keras.models import load_model
-import numpy as np
-import cv2
-from keras.models import load_model
-
-
-class Gender:
-	def __init__(self, model= "pre_trained_gn/gender_detection.model"):
-		self.model = load_model(model)
-		self.classes = ['man', 'woman']
-
-	def f_detector(self, frame, bbox):
-		if frame is None:
-			print("Could not read input image")
-		gender = list()
-		info = {}
-		if np.shape(bbox)[0] is 0 :
-			info = {"person": 0, "man": 0, "woman": 0}
-			return info
-		else:
-			for idx, f in enumerate(bbox):
-				(startx, starty) = f[0], f[1]
-				(endx, endy) = f[2], f[3],
-				conf = self.model.predict(
-					self.crop(frame, startx, starty, endx, endy))[0]
-				idx = np.argmax(conf)
-				label = self.classes[idx]
-				gender.append(label)
-				info = {"person": np.shape(bbox)[0], "man": gender.count("man"), "woman": gender.count("woman")}
-			return info, gender
-
-	def m_detector(self, img):
-		if img is None:
-			print("Could not read input image")
-		face_crop = cv2.resize(img, (96, 96))
-		face_crop = face_crop.astype("float") / 255.0
-		face_crop = img_to_array(face_crop)
-		face_crop = np.expand_dims(face_crop, axis=0)
-		conf = self.model.predict(face_crop)[0]
-		idx = np.argmax(conf)
-		label = self.classes[idx]
-		return label
-
-
-	def crop (self, frame, startx, starty, endx, endy):
-		face_crop = np.copy(frame[int(np.rint(starty)):int(np.rint(endy)), int(np.rint(startx)):int(np.rint(endx))])
-		face_crop = cv2.resize(face_crop, (96, 96))
-		face_crop = face_crop.astype("float") / 255.0
-		face_crop = img_to_array(face_crop)
-		face_crop = np.expand_dims(face_crop, axis=0)
-		return face_crop

src/model_engineering.py

@@ -7,7 +7,6 @@
 import numpy as np
 import tensorflow as tf
 from tensorflow.python.platform import gfile
-from gender_detection import Gender
 from knn import KNN
 
 
@@ -115,7 +114,6 @@ def encode(self, images):
         """
         if not self.initialized:
             self.initialized = self.initialize()
-
         feed_dict = {self.imgs_ph: images, self.phase_train_ph: False}
         emb_array = self.session.run(self.embs_ph, feed_dict=feed_dict)
         return emb_array