def manual_make_test(self):
training_data = {}
blind_test = {}
a = FaceInPictureClassifier()
a.get_listing_ids_from_file("../cached/active_sundress_listing_ids.txt")
cv.NamedWindow("Contains a face? (y/n)")
print "Making a manual test set for recognizing faces."
print "Please press y if a picture contains a full face, n otherwise."
print "To quit, press q."
for i in range(0,199):
# print a.listingIds[i]
if i < 100:
training_data[int(a.listingIds[i])] = {}
else:
blind_test[int(a.listingIds[i])] = {}
a.imgGrabber.get_listing_image(a.listingIds[i])
cv.ShowImage("Contains a face? (y/n)", a.imgGrabber.cvImage)
c = 0
print "Processing image %d of 200" % (i + 1)
while(1):
c = cv.WaitKey(0)
if i < 100:
if c == 121:
training_data[int(a.listingIds[i])]['face'] = True
print "Previous image had face."
break
if c == 110:
training_data[int(a.listingIds[i])]['face'] = False
print "Previous image had no face."
break
else:
if c == 121:
blind_test[int(a.listingIds[i])]['face'] = True
print "Previous image had face."
break
if c == 110:
blind_test[int(a.listingIds[i])]['face'] = False
print "Previous image had no face."
break
if c == 113:
print "received kill command, exiting program."
return 0
pickle.dump(training_data, open("testing/training_data.p", "wb"))
pickle.dump(blind_test, open("testing/blind_test.p", "wb"))
def __init__(self):
#add in classifiers here
self.FIPC = FaceInPictureClassifier()
self.classifiers = ["haarcascade_eyes_22_5.xml",
"haarcascade_eyes_45_11.xml",
"haarcascade_eye.xml",
"haarcascade_frontal_eyes.xml",
"haarcascade_frontalface_alt2.xml",
"haarcascade_frontalface_alt_tree.xml",
"haarcascade_frontalface_alt.xml",
"haarcascade_frontalface_default.xml",
"haarcascade_fullbody.xml",
"haarcascade_head_and_shoulders.xml",
"haarcascade_profileface.xml"
]
class ClassifierTest():
def __init__(self):
#add in classifiers here
self.FIPC = FaceInPictureClassifier()
self.classifiers = ["haarcascade_eyes_22_5.xml",
"haarcascade_eyes_45_11.xml",
"haarcascade_eye.xml",
"haarcascade_frontal_eyes.xml",
"haarcascade_frontalface_alt2.xml",
"haarcascade_frontalface_alt_tree.xml",
"haarcascade_frontalface_alt.xml",
"haarcascade_frontalface_default.xml",
"haarcascade_fullbody.xml",
"haarcascade_head_and_shoulders.xml",
"haarcascade_profileface.xml"
]
def manual_make_test(self):
training_data = {}
blind_test = {}
a = FaceInPictureClassifier()
a.get_listing_ids_from_file("../cached/active_sundress_listing_ids.txt")
cv.NamedWindow("Contains a face? (y/n)")
print "Making a manual test set for recognizing faces."
print "Please press y if a picture contains a full face, n otherwise."
print "To quit, press q."
for i in range(0,199):
# print a.listingIds[i]
if i < 100:
training_data[int(a.listingIds[i])] = {}
else:
blind_test[int(a.listingIds[i])] = {}
a.imgGrabber.get_listing_image(a.listingIds[i])
cv.ShowImage("Contains a face? (y/n)", a.imgGrabber.cvImage)
c = 0
print "Processing image %d of 200" % (i + 1)
while(1):
c = cv.WaitKey(0)
if i < 100:
if c == 121:
training_data[int(a.listingIds[i])]['face'] = True
print "Previous image had face."
break
if c == 110:
training_data[int(a.listingIds[i])]['face'] = False
print "Previous image had no face."
break
else:
if c == 121:
blind_test[int(a.listingIds[i])]['face'] = True
print "Previous image had face."
break
if c == 110:
blind_test[int(a.listingIds[i])]['face'] = False
print "Previous image had no face."
break
if c == 113:
print "received kill command, exiting program."
return 0
pickle.dump(training_data, open("testing/training_data.p", "wb"))
pickle.dump(blind_test, open("testing/blind_test.p", "wb"))
def run_tests(self):
training_data_faces = 0
detection_frontal = 0
false_positive_frontal = 0
detection_profile = 0
false_positive_profile = 0
detection_both = 0
detection_either = 0
false_positive_both = 0
false_positive_either = 0
training_data = pickle.load(open("testing/blind_test.p", "rb"))
blind_test = pickle.load(open("testing/blind_test.p", "rb"))
frontal_face_results = pickle.load(open("results/frontalface_alt.p", "rb"))
profile_face_results = pickle.load(open("results/frontalface_alt2.p", "rb"))
for listing in training_data:
realFace = training_data[listing]['face']
frontFace = frontal_face_results[int(listing)]['face']
profileFace = profile_face_results[int(listing)]['face']
if realFace:
training_data_faces = training_data_faces + 1
if realFace and frontFace:
detection_frontal = detection_frontal + 1
if realFace and profileFace:
detection_profile = detection_profile + 1
if frontFace and profileFace:
if realFace:
detection_both = detection_both + 1
else:
false_positive_both = false_positive_both + 1
if frontFace or profileFace:
if realFace:
detection_either = detection_either + 1
else:
false_positive_either = false_positive_either + 1
if frontFace and not realFace:
false_positive_frontal = false_positive_frontal + 1
if profileFace and not realFace:
false_positive_profile = false_positive_profile + 1
print("Test results:")
print("Out of a possible %d faces." % training_data_faces)
print("Frontal Face Classifier: \n%d detected, %d false positives" % (detection_frontal, false_positive_frontal))
print("Profile Face Classifier: \n%d detected, %d false positives" % (detection_profile, false_positive_profile))
print("Both Face Classifiers: \n%d detected, %d false positives" % (detection_both, false_positive_both))
print("Either Face Classifier: \n%d detected, %d false positives" % (detection_either, false_positive_either))
def make_new_test_data(self):
training_data = pickle.load(open("testing/training_data.p", "rb"))
blind_test_data = pickle.load(open("testing/blind_test.p", "rb"))
trainingListIds = [int(x) for x in training_data]
blindTestIds = [int(x) for x in blind_test_data]
trainingListIds += blindTestIds
self.FIPC.set_listing_ids(trainingListIds)
for classifier in self.classifiers:
print("Beginning classification for %s" % classifier[12:-4])
self.FIPC.check_for_faces("classifiers/" + classifier)
pickle.dump(self.FIPC.listingStatistics, open("results/" + classifier[12:-4]+ ".p", "wb"))
def show_results(self):
training_data = pickle.load(open("testing/blind_test.p", "rb"))
self.classifierData = []
trainingListIds = [int(x) for x in training_data]
#single results:
for classifier in self.classifiers:
print("Classifier: %s returned the following:" % classifier)
temp = pickle.load(open("results/" + classifier[12:-4] + ".p", "rb"))
classifierResults = self.single_test(training_data, temp)
classifierData.append[temp]
print("Correct: %d" % classifierResults['Correct'])
print("Detections: %d" % classifierResults['Detections'])
print("False Positives: %d" % classifierResults['False Positives'])
#mixed results:
def single_test(self, training_data, test_data):
testResults = {'Correct': 0,
'Detections': 0,
'False Positives': 0}
for listingId in training_data:
real = training_data[listingId]['face']
test = test_data[int(listingId)]['face']
if real == test:
testResults['Correct'] += 1
if real and test:
testResults['Detections'] += 1
if not real and test:
testResults['False Positives'] += 1
return testResults
def merge_two_tests_and(self, test_data1, test_data2):
mergedTestData = {}
for listing in test_data1:
mergedTestData[listing] = {}
mergedTestData[listing]['face'] = test_data1[listing]['face'] and test_data2[listing]['face']
return mergedTestData
