├── README.md
├── test.py
└── hog.py


/README.md:
--------------------------------------------------------------------------------
1 | # Face_Detection_HOG
2 | Face Detection using HOG and SVM
3 | The training file for the data is hog.py
4 | The testing file is test.py......
5 | 
6 | Now to train on more data... Update the file hog.py accordingly.
7 | 


--------------------------------------------------------------------------------
/test.py:
--------------------------------------------------------------------------------
 1 | import cv2
 2 | from sklearn.externals import joblib
 3 | from skimage.feature import hog
 4 | import numpy as np
 5 | 
 6 | h = 128
 7 | w = 64
 8 | 
 9 | # Load the classifier
10 | clf = joblib.load("hog.pkl")
11 | 
12 | # Read the input image 
13 | im = cv2.imread("image1048.tif")
14 | 
15 | img=cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
16 | img=cv2.resize(img, (w, h), interpolation=cv2.INTER_CUBIC)
17 | 
18 | roi_hog_fd = hog(img, orientations=105, pixels_per_cell=(14, 14), cells_per_block=(1, 1), visualise=False)
19 | 
20 | nbr = clf.predict(np.array([roi_hog_fd], 'float64'))
21 | 
22 | print nbr
23 | 


--------------------------------------------------------------------------------
/hog.py:
--------------------------------------------------------------------------------
 1 | import cv2
 2 | import numpy as np
 3 | import os
 4 | from sklearn.svm import SVC
 5 | import matplotlib.pyplot as plt
 6 | from sklearn.externals import joblib
 7 | #import scikit-image.skimage.data
 8 | #, color, exposure
 9 | 
10 | 
11 | array=np.array([])	#empty array for storing all the features
12 | h = 128			#height of the image
13 | w = 64			#width of the image
14 | hog=cv2.HOGDescriptor()
15 | img=cv2.imread('image1048.tif',0)
16 | img1=cv2.imread('image1049.tif',0)
17 | 
18 | img=cv2.resize(img, (w, h), interpolation=cv2.INTER_CUBIC)	#resize images
19 | img1=cv2.resize(img1, (w, h), interpolation=cv2.INTER_CUBIC)
20 | for i in range(0,2):
21 | 	if i==0:
22 | 		h=hog.compute(img,winStride=(64,128),padding=(0,0))	#storing HOG features as column vector
23 | 		#hog_image_rescaled = exposure.rescale_intensity(h, in_range=(0, 0.02))
24 | 		#plt.figure(1, figsize=(3, 3))
25 | 		#plt.imshow(h,cmap=plt.cm.gray)
26 | 		#plt.show()
27 | 		#print len(h)
28 | 		h_trans=h.transpose()	#transposing the column vector
29 | 	
30 | 		array=np.vstack(h_trans)	#appending it to the array
31 | 		print "HOG features of label 1"
32 | 		print array
33 | 		
34 | 	else:
35 | 		h=hog.compute(img1,winStride=(64,128),padding=(0,0))	#storing HOG features as column vector
36 | 		
37 | 		#hog_image_rescaled = exposure.rescale_intensity(h, in_range=(0, 0.02))
38 | 		#plt.figure(1, figsize=(3, 3))
39 | 		#plt.imshow(h,cmap=plt.cm.gray)
40 | 		#plt.show()
41 | 		#print len(h)
42 | 		h_trans=h.transpose()	#transposing the column vector
43 | 	
44 | 		array=np.vstack((array,h_trans))	#appending it to the array
45 | 		print "HOG features of label 1 & 2"
46 | 		print (array)	
47 | 
48 | #print (array.shape)
49 | 
50 | 
51 | label=[1,4]
52 | 
53 | clf=SVC(gamma=0.001,C=10)
54 | 
55 | clf.fit(array,label)
56 | 
57 | #ypred = clf.predict()
58 | joblib.dump(clf, "hog.pkl", compress=3)
59 | 


--------------------------------------------------------------------------------