def cropFaces(self, cascade=None):
"""
**SUMMARY**
This function helps in cropping a certain object in all of the images
by using the provided haar cascade. A haar classifier is implemented
and images are cropped. This function also supports multiple faces in
one single image.
**PARAMETERS**
cascade - haar cascade string
**RETURNS**
ImageSet
**EXAMPLES**
>>> imgs = ImageSet("some_directory/")
>>> cascade = "face.xml"
>>> faces = imgs.cropFaces(cascade)
>>> faces.show()
"""
if not cascade:
cascade = CASCADE_PATH
else:
if not os.path.isfile(cascade):
warnings.warn(
"The provided cascade does not exist. Using default cascade"
)
cascade = CASCADE_PATH
classifier = cv2.CascadeClassifier(cascade)
gray = [cv2.cvtColor(img, cv2.cv.CV_BGR2GRAY) for img in self]
objects = [
classifier.detectMultiScale(img,
scaleFactor=1.1,
minNeighbors=3,
minSize=(10, 10),
flags=cv2.cv.CV_HAAR_SCALE_IMAGE)
for img in gray
]
imgs = [[img[ob[1]:ob[1] + ob[3], ob[0]:ob[0] + ob[2]] for ob in obj]
for img, obj in zip(self, objects)
if isinstance(obj, np.ndarray)]
imgs = concatenate(*imgs)
return ImageSet(imgs=imgs)
def __init__(self):
"""
Create a Face Recognizer Class using Fisher Face Recognizer. Uses
OpenCV's FaceRecognizer class. Currently supports Fisher Faces.
"""
self.supported = True
self.model = None
self.train_imgs = None
self.train_labels = None
self.csvfiles = []
self.imageSize = None
self.labels_dict = {}
self.labels_set = []
self.int_labels = []
self.labels_dict_rev = {}
if not hasattr(cv2, 'createFisherFaceRecognizer'):
self.supported = False
warnings.warn("Returning None. OpenCV >= 2.4.4 required.")
return
self.model = cv2.createFisherFaceRecognizer()
def __init__(self, directory="", imgs=None):
if os.path.isfile(directory):
img = cv2.imread(directory)
self.append(img)
return
if imgs:
self.extend(imgs)
return
try:
imagefiles = os.listdir(directory)
except OSError as error:
print "OS Error({0}): {1}" .format(error.errno, error.strerror)
warnings.warn("encountered the above mentioned error. Returning Empty list.")
return
for imagefile in imagefiles:
filename = os.path.join(directory, imagefile)
img = cv2.imread(filename)
if isinstance(img, np.ndarray):
self.append(img)
def __init__(self):
"""
Create a Face Recognizer Class using Fisher Face Recognizer. Uses
OpenCV's FaceRecognizer class. Currently supports Fisher Faces.
"""
self.supported = True
self.model = None
self.train_imgs = None
self.train_labels = None
self.csvfiles = []
self.imageSize = None
self.labels_dict = {}
self.labels_set = []
self.int_labels = []
self.labels_dict_rev = {}
if not hasattr(cv2, 'createFisherFaceRecognizer'):
self.supported = False
warnings.warn("Returning None. OpenCV >= 2.4.4 required.")
return
self.model = cv2.createFisherFaceRecognizer()
def __init__(self, directory="", imgs=None):
if os.path.isfile(directory):
img = cv2.imread(directory)
self.append(img)
return
if imgs:
self.extend(imgs)
return
try:
imagefiles = os.listdir(directory)
except OSError as error:
print "OS Error({0}): {1}".format(error.errno, error.strerror)
warnings.warn(
"encountered the above mentioned error. Returning Empty list.")
return
for imagefile in imagefiles:
filename = os.path.join(directory, imagefile)
img = cv2.imread(filename)
if isinstance(img, np.ndarray):
self.append(img)
def load(self, filename):
"""
**SUMMARY**
Load the trainging data.
**PARAMETERS**
* *filename* - File where you want to load the data from.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> f.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
self.model.load(filename)
loadfile = open(filename, "r")
for line in loadfile.readlines():
if "cols" in line:
match = re.search("(?<=\>)\w+", line)
tsize = int(match.group(0))
break
loadfile.close()
w = int(tsize ** 0.5)
h = tsize / w
while(w * h != tsize):
w += 1
h = tsize / w
self.imageSize = (w, h)
def load(self, filename):
"""
**SUMMARY**
Load the trainging data.
**PARAMETERS**
* *filename* - File where you want to load the data from.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> f.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
self.model.load(filename)
loadfile = open(filename, "r")
for line in loadfile.readlines():
if "cols" in line:
match = re.search("(?<=\>)\w+", line)
tsize = int(match.group(0))
break
loadfile.close()
w = int(tsize**0.5)
h = tsize / w
while (w * h != tsize):
w += 1
h = tsize / w
self.imageSize = (w, h)
def save(self, filename):
"""
**SUMMARY**
Save the trainging data.
**PARAMETERS**
* *filename* - File where you want to save the data.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
#Save New Fisher Training Data
>>> f.save("new_trainingdata.xml")
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
self.model.save(filename)
def save(self, filename):
"""
**SUMMARY**
Save the trainging data.
**PARAMETERS**
* *filename* - File where you want to save the data.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
#Save New Fisher Training Data
>>> f.save("new_trainingdata.xml")
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
self.model.save(filename)
def cropFaces(self, cascade=None):
"""
**SUMMARY**
This function helps in cropping a certain object in all of the images
by using the provided haar cascade. A haar classifier is implemented
and images are cropped. This function also supports multiple faces in
one single image.
**PARAMETERS**
cascade - haar cascade string
**RETURNS**
ImageSet
**EXAMPLES**
>>> imgs = ImageSet("some_directory/")
>>> cascade = "face.xml"
>>> faces = imgs.cropFaces(cascade)
>>> faces.show()
"""
if not cascade:
cascade = CASCADE_PATH
else:
if not os.path.isfile(cascade):
warnings.warn("The provided cascade does not exist. Using default cascade")
cascade = CASCADE_PATH
classifier = cv2.CascadeClassifier(cascade)
gray = [cv2.cvtColor(img, cv2.cv.CV_BGR2GRAY) for img in self]
objects = [classifier.detectMultiScale(img, scaleFactor=1.1, minNeighbors=3, minSize=(10, 10), flags = cv2.cv.CV_HAAR_SCALE_IMAGE) for img in gray]
imgs = [[img[ob[1]:ob[1]+ob[3], ob[0]:ob[0]+ob[2]] for ob in obj] for img, obj in zip(self, objects) if isinstance(obj, np.ndarray)]
imgs = concatenate(*imgs)
return ImageSet(imgs=imgs)
def __init__(self, label, imageset):
if not isinstance(imageset, collections.Iterable):
warnings.warn("The provided ImageSet is not a list")
return None
labels = [label]*len(imageset)
self.extend(labels)
def train(self, images=None, labels=None, csvfile=None, delimiter=";"):
"""
**SUMMARY**
Train the face recognizer with images and labels.
**PARAMETERS**
* *images* - A list of Images or ImageSet. All the images must be of
same size.
* *labels* - A list of labels(int) corresponding to the image in
images.
There must be at least two different labels.
* *csvfile* - You can also provide a csv file with image filenames
and labels instead of providing labels and images
separately.
* *delimiter* - The delimiter used in csv files.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
Save Fisher Training Data
>>> f.save("trainingdata.xml")
Load Fisher Training Data and directly use without trainging
>>> f1 = FaceRecognizer()
>>> f1.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f1.predict(imgs)
Use CSV files for training
>>> f = FaceRecognizer()
>>> f.train(csvfile="CSV_file_name", delimiter=";")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if csvfile:
images = []
labels = []
import csv
try:
f = open(csvfile, "rb")
except IOError:
warnings.warn("No such file found. Training not initiated")
return None
self.csvfiles.append(csvfile)
filereader = csv.reader(f, delimiter=delimiter)
for row in filereader:
images.append(Image(row[0]))
labels.append(row[1])
if isinstance(labels, type(None)):
warnings.warn("Labels not provided. Training not inititated.")
return None
self.labels_set = list(set(labels))
i = 0
for label in self.labels_set:
self.labels_dict.update({label: i})
self.labels_dict_rev.update({i: label})
i += 1
if len(self.labels_set) < 2:
warnings.warn("At least two classes/labels are required"
"for training. Training not inititated.")
return None
if len(images) != len(labels):
warnings.warn("Mismatch in number of labels and number of"
"training images. Training not initiated.")
return None
self.imageSize = images[0].shape[:2]
h, w = self.imageSize
images = [img if img.shape[:2] == self.imageSize
else cv2.resize(img, (w, h)) for img in images]
self.int_labels = [self.labels_dict[key] for key in labels]
self.train_labels = labels
labels = np.array(self.int_labels)
self.train_imgs = images
cv2imgs = [cv2.cvtColor(img, cv2.cv.CV_BGR2GRAY) for img in images]
self.model.train(cv2imgs, labels)
def predict(self, imgs):
"""
**SUMMARY**
Predict the class of the image using trained face recognizer.
**PARAMETERS**
* *image* - Image.The images must be of the same size as provided
in training.
**RETURNS**
* *label* - Class of the image which it belongs to.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
Save Fisher Training Data
>>> f.save("trainingdata.xml")
Load Fisher Training Data and directly use without trainging
>>> f1 = FaceRecognizer()
>>> f1.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f1.predict(imgs)
Use CSV files for training
>>> f = FaceRecognizer()
>>> f.train(csvfile="CSV_file_name", delimiter=";")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
h, w = self.imageSize
images = [img if img.shape[:2] == self.imageSize
else cv2.resize(img, (w, h)) for img in imgs]
if isinstance(imgs, np.ndarray):
if imgs.shape[:2] != self.imageSize:
image = cv2.resize(imgs, (w, h))
cv2img = cv2.cvtColor(image, cv2.cv.CV_BGR2GRAY)
label, confidence = self.model.predict(cv2img)
retLabel = self.labels_dict_rev.get(label)
if not retLabel:
retLabel = label
return (retLabel, confidence)
retVal = []
for image in images:
cv2img = cv2.cvtColor(image, cv2.cv.CV_BGR2GRAY)
label, confidence = self.model.predict(cv2img)
retLabel = self.labels_dict_rev.get(label)
if not retLabel:
retLabel = label
retVal.append((retLabel, confidence))
return retVal
def __init__(self, label, imageset):
if not isinstance(imageset, collections.Iterable):
warnings.warn("The provided ImageSet is not a list")
return None
labels = [label] * len(imageset)
self.extend(labels)
def train(self, images=None, labels=None, csvfile=None, delimiter=";"):
"""
**SUMMARY**
Train the face recognizer with images and labels.
**PARAMETERS**
* *images* - A list of Images or ImageSet. All the images must be of
same size.
* *labels* - A list of labels(int) corresponding to the image in
images.
There must be at least two different labels.
* *csvfile* - You can also provide a csv file with image filenames
and labels instead of providing labels and images
separately.
* *delimiter* - The delimiter used in csv files.
**RETURNS**
Nothing. None.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
Save Fisher Training Data
>>> f.save("trainingdata.xml")
Load Fisher Training Data and directly use without trainging
>>> f1 = FaceRecognizer()
>>> f1.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f1.predict(imgs)
Use CSV files for training
>>> f = FaceRecognizer()
>>> f.train(csvfile="CSV_file_name", delimiter=";")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if csvfile:
images = []
labels = []
import csv
try:
f = open(csvfile, "rb")
except IOError:
warnings.warn("No such file found. Training not initiated")
return None
self.csvfiles.append(csvfile)
filereader = csv.reader(f, delimiter=delimiter)
for row in filereader:
images.append(Image(row[0]))
labels.append(row[1])
if isinstance(labels, type(None)):
warnings.warn("Labels not provided. Training not inititated.")
return None
self.labels_set = list(set(labels))
i = 0
for label in self.labels_set:
self.labels_dict.update({label: i})
self.labels_dict_rev.update({i: label})
i += 1
if len(self.labels_set) < 2:
warnings.warn("At least two classes/labels are required"
"for training. Training not inititated.")
return None
if len(images) != len(labels):
warnings.warn("Mismatch in number of labels and number of"
"training images. Training not initiated.")
return None
self.imageSize = images[0].shape[:2]
h, w = self.imageSize
images = [
img if img.shape[:2] == self.imageSize else cv2.resize(
img, (w, h)) for img in images
]
self.int_labels = [self.labels_dict[key] for key in labels]
self.train_labels = labels
labels = np.array(self.int_labels)
self.train_imgs = images
cv2imgs = [cv2.cvtColor(img, cv2.cv.CV_BGR2GRAY) for img in images]
self.model.train(cv2imgs, labels)
def predict(self, imgs):
"""
**SUMMARY**
Predict the class of the image using trained face recognizer.
**PARAMETERS**
* *image* - Image.The images must be of the same size as provided
in training.
**RETURNS**
* *label* - Class of the image which it belongs to.
**EXAMPLES**
>>> f = FaceRecognizer()
>>> imgs1 = ImageSet(path/to/images_of_type1)
>>> labels1 = LabelSet("type1", imgs1)
>>> imgs2 = ImageSet(path/to/images_of_type2)
>>> labels2 = LabelSet("type2", imgs2)
>>> imgs3 = ImageSet(path/to/images_of_type3)
>>> labels3 = LabelSet("type3", imgs3)
>>> imgs = concatenate(imgs1, imgs2, imgs3)
>>> labels = concatenate(labels1, labels2, labels3)
>>> f.train(imgs, labels)
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
Save Fisher Training Data
>>> f.save("trainingdata.xml")
Load Fisher Training Data and directly use without trainging
>>> f1 = FaceRecognizer()
>>> f1.load("trainingdata.xml")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f1.predict(imgs)
Use CSV files for training
>>> f = FaceRecognizer()
>>> f.train(csvfile="CSV_file_name", delimiter=";")
>>> imgs = ImageSet("path/to/testing_images")
>>> print f.predict(imgs)
"""
if not self.supported:
warnings.warn("Fisher Recognizer is supported by OpenCV >= 2.4.4")
return None
h, w = self.imageSize
images = [
img if img.shape[:2] == self.imageSize else cv2.resize(
img, (w, h)) for img in imgs
]
if isinstance(imgs, np.ndarray):
if imgs.shape[:2] != self.imageSize:
image = cv2.resize(imgs, (w, h))
cv2img = cv2.cvtColor(image, cv2.cv.CV_BGR2GRAY)
label, confidence = self.model.predict(cv2img)
retLabel = self.labels_dict_rev.get(label)
if not retLabel:
retLabel = label
return (retLabel, confidence)
retVal = []
for image in images:
cv2img = cv2.cvtColor(image, cv2.cv.CV_BGR2GRAY)
label, confidence = self.model.predict(cv2img)
retLabel = self.labels_dict_rev.get(label)
if not retLabel:
retLabel = label
retVal.append((retLabel, confidence))
return retVal