def test_rgba_int_array_facial_features(self):
test_face = os.path.normpath(os.path.join(DIR, "data/48by48rgba.png"))
response = facial_features(test_face)
self.assertTrue(isinstance(response, list))
self.assertEqual(len(response), 48)
self.check_range(response)
def test_good_facial_features(self):
test_face = np.random.rand(48,48).tolist()
response = facial_features(test_face)
self.assertTrue(isinstance(response, list))
self.assertEqual(len(response), 48)
self.check_range(response)
def test_good_facial_features(self):
test_face = np.random.rand(48, 48).tolist()
response = facial_features(test_face)
self.assertTrue(isinstance(response, list))
self.assertEqual(len(response), 48)
self.check_range(response)
def test_good_int_array_facial_features(self):
fer_set = set(['Angry', 'Sad', 'Neutral', 'Surprise', 'Fear', 'Happy'])
test_face = os.path.normpath(os.path.join(DIR, "data/48by48.png"))
response = facial_features(test_face)
self.assertTrue(isinstance(response, list))
self.assertEqual(len(response), 48)
self.check_range(response)
def _grab_one(self, local_img_path, search_term):
try:
img = Image.open(local_img_path)
self.captured_data.append( {'celeb': search_term,
'face_feats': indicoio.facial_features(img),
'face_corners': self._get_single_face_corners(img),
'local_img_path': local_img_path
})
except:
print('failed to grab facial feats for ' + local_img_path)
self.failed_to_capture.append((local_img_path, search_term))
def _grab_one(self, local_img_path, search_term):
try:
img = Image.open(local_img_path)
self.captured_data.append({
'celeb':
search_term,
'face_feats':
indicoio.facial_features(img),
'face_corners':
self._get_single_face_corners(img),
'local_img_path':
local_img_path
})
except:
print('failed to grab facial feats for ' + local_img_path)
self.failed_to_capture.append((local_img_path, search_term))
def recordFaceData(directory, nickname):
allImages = []
count = 0
for file in os.listdir(directory):
if file.endswith('.jpg'):
allImages.append(file)
currentRecordedFeatures = {}
currentRecordedFeatures['names'] = []
currentRecordedFeatures['features'] = []
print 'Caching...'
for image in allImages:
print 'Collecting info ... ', count
count += 1
face = indicoio.facial_localization(directory + '/' + image)
if len(face) > 0:
face = face[0]
img = cv2.imread(directory + '/' + image)
faceImg = img[
face['top_left_corner'][1] : face['bottom_right_corner'][1],
face['top_left_corner'][0] : face['bottom_right_corner'][0]
]
facialFeatures = indicoio.facial_features(faceImg)
if (count % 50) == 0:
cv2.imwrite('images/faces/' + str(int(time.time())) + '.jpg', faceImg)
currentRecordedFeatures['names'].append(nickname)
currentRecordedFeatures['features'].append(facialFeatures)
np.save(featuresCacheFile + nickname + '.npy', [currentRecordedFeatures])
print 'Done'
return currentRecordedFeatures
def test_batch_facial_features(self):
test_data = [os.path.normpath(os.path.join(DIR, "data/48by48.png"))]
response = facial_features(test_data, api_key=self.api_key)
self.assertTrue(isinstance(response, list))
self.assertTrue(isinstance(response[0], list))
self.assertEqual(len(response[0]), 48)
def test_good_facial_features(self):
test_face = np.linspace(0,50,48*48).reshape(48,48).tolist()
response = facial_features(test_face)
self.assertTrue(isinstance(response, list))
self.assertEqual(len(response), 48)
def identifyFaces(imgPath, testing=False):
faces = extractFaces(imgPath)
dataMisha = np.load('facialFeatures/facesMisha.npy')[0]
dataMihai = np.load('facialFeatures/facesMihai.npy')[0]
dataPavel = np.load('facialFeatures/facesPavel.npy')[0]
dataRishab = np.load('facialFeatures/facesRishab.npy')[0]
trainingNames = dataMisha['names'] + \
dataMihai['names'] + \
dataPavel['names'] + \
dataRishab['names']
trainingFeatures = dataMisha['features'] + \
dataMihai['features'] + \
dataPavel['features'] + \
dataRishab['features']
n_trainingNames = lb.fit_transform(trainingNames)
# clf = SGDClassifier(loss="modified_huber", penalty="l1")
# clf = SGDClassifier(loss="modified_huber", penalty="l1", epsilon=0.0001)
# clf = NearestCentroid()
clf = KNeighborsClassifier(n_neighbors=5, weights='uniform', algorithm='auto', leaf_size=30, p=2, metric='minkowski', metric_params=None)
clf.fit(trainingFeatures, n_trainingNames)
for path in faces:
print lb.inverse_transform(clf.predict(indicoio.facial_features(path)))
if testing == True:
predictedNames = []
trainingNames = dataMisha['names'][0:650] + \
dataMihai['names'][0:650] + \
dataPavel['names'][0:650] + \
dataRishab['names'][0:650]
trainingFeatures = dataMisha['features'][0:650] + \
dataMihai['features'][0:650] + \
dataPavel['features'][0:650] + \
dataRishab['features'][0:650]
testNames = dataMisha['names'][651:701] + \
dataMihai['names'][651:701] + \
dataPavel['names'][651:701] + \
dataRishab['names'][651:701]
testFeatures = dataMisha['features'][651:701] + \
dataMihai['features'][651:701] + \
dataPavel['features'][651:701] + \
dataRishab['features'][651:701]
correctCount = 0
count = 0
for x in range(0,len(testFeatures)):
predictedName = clf.predict(testFeatures[x])
predictedNames.append(predictedName)
predictedNames = lb.inverse_transform(predictedNames)
for x in range(0,len(predictedNames)):
print count , ' *** ', predictedNames[x], ' - ', testNames[x]
count += 1
if predictedNames[x] == testNames[x]:
correctCount += 1
print 'Correct % - ', (correctCount / float(len(predictedNames))) * 100
print correctCount, '/' , len(predictedNames)
def get_images_feats(images_paths):
"returns list of np arrays of feats"
results = indicoio.facial_features(images_paths)
return [np.array(results[i]) for i in xrange(len(results))]
