diff --git a/bob/pad/face/config/preprocessor/video_face_crop.py b/bob/pad/face/config/preprocessor/video_face_crop.py
new file mode 100644
index 0000000000000000000000000000000000000000..ad7f963b0fdb8f92319c02ad8210c899a44dbccc
--- /dev/null
+++ b/bob/pad/face/config/preprocessor/video_face_crop.py
@@ -0,0 +1,31 @@
+#!/usr/bin/env python2
+# -*- coding: utf-8 -*-
+
+from bob.pad.face.preprocessor import VideoFaceCrop
+
+#=======================================================================================
+# Define instances here:
+
+CROPPED_IMAGE_SIZE = (64, 64) # The size of the resulting face
+CROPPED_POSITIONS = {'topleft' : (0,0) , 'bottomright' : CROPPED_IMAGE_SIZE}
+FIXED_POSITIONS = None
+MASK_SIGMA = None # The sigma for random values areas outside image
+MASK_NEIGHBORS = 5 # The number of neighbors to consider while extrapolating
+MASK_SEED = None # The seed for generating random values during extrapolation
+CHECK_FACE_SIZE_FLAG = True # Check the size of the face
+MIN_FACE_SIZE = 50
+USE_LOCAL_CROPPER_FLAG = True # Use the local face cropping class (identical to Ivana's paper)
+RGB_OUTPUT_FLAG = True # Return RGB cropped face using local cropper
+DETECT_FACES_FLAG = True # find annotations locally replacing the database annotations
+
+preprocessor_rgb_face_detect = VideoFaceCrop(cropped_image_size = CROPPED_IMAGE_SIZE,
+ cropped_positions = CROPPED_POSITIONS,
+ fixed_positions = FIXED_POSITIONS,
+ mask_sigma = MASK_SIGMA,
+ mask_neighbors = MASK_NEIGHBORS,
+ mask_seed = None,
+ check_face_size_flag = CHECK_FACE_SIZE_FLAG,
+ min_face_size = MIN_FACE_SIZE,
+ use_local_cropper_flag = USE_LOCAL_CROPPER_FLAG,
+ rgb_output_flag = RGB_OUTPUT_FLAG,
+ detect_faces_flag = DETECT_FACES_FLAG)
diff --git a/bob/pad/face/preprocessor/VideoFaceCrop.py b/bob/pad/face/preprocessor/VideoFaceCrop.py
index fd9756766aea41ad1de85a1f9ae398ac2809f7f3..97024f8daa0b7f282206025b4e7643020f24bdb1 100644
--- a/bob/pad/face/preprocessor/VideoFaceCrop.py
+++ b/bob/pad/face/preprocessor/VideoFaceCrop.py
@@ -18,6 +18,8 @@ import numpy as np
from bob.pad.face.preprocessor.ImageFaceCrop import ImageFaceCrop
+from ..utils.face_detection_utils import detect_faces_in_video
+
#==============================================================================
# Main body:
@@ -76,6 +78,12 @@ class VideoFaceCrop(Preprocessor, object):
returned. This flag is only valid when ``use_local_cropper_flag = True``.
Default: ``False``.
+ ``detect_faces_flag`` : :py:class:`bool`
+ If set to ``True`` the facial annotations will be generated using
+ face detection. Otherwise, annotations of the database are used for
+ cropping.
+ Default: ``False``.
+
``kwargs``
Remaining keyword parameters passed to the Base constructor, such as ``color_channel`` or ``dtype``.
"""
@@ -92,6 +100,7 @@ class VideoFaceCrop(Preprocessor, object):
min_face_size = 50,
use_local_cropper_flag = False,
rgb_output_flag = False,
+ detect_faces_flag = False,
**kwargs):
super(VideoFaceCrop, self).__init__(cropped_image_size = cropped_image_size,
@@ -104,6 +113,7 @@ class VideoFaceCrop(Preprocessor, object):
min_face_size = min_face_size,
use_local_cropper_flag = use_local_cropper_flag,
rgb_output_flag = rgb_output_flag,
+ detect_faces_flag = detect_faces_flag,
**kwargs)
self.cropped_image_size = cropped_image_size
@@ -116,6 +126,7 @@ class VideoFaceCrop(Preprocessor, object):
self.min_face_size = min_face_size
self.use_local_cropper_flag = use_local_cropper_flag
self.rgb_output_flag = rgb_output_flag
+ self.detect_faces_flag = detect_faces_flag
# Save also the data stored in the kwargs:
for (k, v) in kwargs.items():
@@ -263,6 +274,10 @@ class VideoFaceCrop(Preprocessor, object):
Cropped faces stored in the FrameContainer.
"""
+ if self.detect_faces_flag:
+
+ annotations = detect_faces_in_video(frames)
+
if len(frames) != len(annotations): # if some annotations are missing
## Select only annotated frames:
diff --git a/bob/pad/face/test/test.py b/bob/pad/face/test/test.py
index 160fbdc9600d8ebf60269ea99e33093e90a51b1c..06dcf820fc7a6c2ecdaac9b0f33dc430e0947c8a 100644
--- a/bob/pad/face/test/test.py
+++ b/bob/pad/face/test/test.py
@@ -29,8 +29,12 @@ from ..extractor import FrameDiffFeatures
from ..extractor import VideoLBPHistogram
+from ..extractor import VideoQualityMeasure
+
from ..algorithm import VideoSvmPadAlgorithm
+from ..algorithm import VideoGmmPadAlgorithm
+
import random
#==============================================================================
@@ -149,6 +153,43 @@ def test_video_face_crop():
assert np.sum(faces[0][1]) == 429158
assert np.sum(faces[-1][1]) == 429158
+ #==========================================================================
+ # test another configuration of the VideoFaceCrop preprocessor:
+
+ CROPPED_IMAGE_SIZE = (64, 64) # The size of the resulting face
+ CROPPED_POSITIONS = {'topleft' : (0,0) , 'bottomright' : CROPPED_IMAGE_SIZE}
+ FIXED_POSITIONS = None
+ MASK_SIGMA = None # The sigma for random values areas outside image
+ MASK_NEIGHBORS = 5 # The number of neighbors to consider while extrapolating
+ MASK_SEED = None # The seed for generating random values during extrapolation
+ CHECK_FACE_SIZE_FLAG = True # Check the size of the face
+ MIN_FACE_SIZE = 50
+ USE_LOCAL_CROPPER_FLAG = True # Use the local face cropping class (identical to Ivana's paper)
+ RGB_OUTPUT_FLAG = True # Return RGB cropped face using local cropper
+ DETECT_FACES_FLAG = True # find annotations locally replacing the database annotations
+
+ preprocessor = VideoFaceCrop(cropped_image_size = CROPPED_IMAGE_SIZE,
+ cropped_positions = CROPPED_POSITIONS,
+ fixed_positions = FIXED_POSITIONS,
+ mask_sigma = MASK_SIGMA,
+ mask_neighbors = MASK_NEIGHBORS,
+ mask_seed = None,
+ check_face_size_flag = CHECK_FACE_SIZE_FLAG,
+ min_face_size = MIN_FACE_SIZE,
+ use_local_cropper_flag = USE_LOCAL_CROPPER_FLAG,
+ rgb_output_flag = RGB_OUTPUT_FLAG,
+ detect_faces_flag = DETECT_FACES_FLAG)
+
+ video, _ = convert_image_to_video_data(image, annotations, 3)
+
+ faces = preprocessor(frames = video, annotations = annotations)
+
+ assert len(faces) == 3
+ assert faces[0][1].shape == (3, 64, 64)
+ assert faces[-1][1].shape == (3, 64, 64)
+ assert np.sum(faces[0][1]) == 1253048
+ assert np.sum(faces[-1][1]) == 1253048
+
#==============================================================================
def test_frame_difference():
@@ -261,6 +302,34 @@ def test_video_lbp_histogram():
assert (lbp_histograms[0][1][-1] - 0.031737773152965658) < 0.000001
+#==============================================================================
+def test_video_quality_measure():
+ """
+ Test VideoQualityMeasure extractor.
+ """
+
+ image = load(datafile('test_image.png', 'bob.pad.face.test'))
+ annotations = {'topleft': (95, 155), 'bottomright': (215, 265)}
+
+ video, annotations = convert_image_to_video_data(image, annotations, 2)
+
+ GALBALLY=True
+ MSU=True
+ DTYPE=None
+
+ extractor = VideoQualityMeasure(galbally=GALBALLY,
+ msu=MSU,
+ dtype=DTYPE)
+
+ features = extractor(video)
+
+ assert len(features) == 2
+ assert len(features[0][1]) == 139
+ assert (features[0][1]==features[-1][1]).all()
+ assert (features[0][1][0] - 2.7748559659812599e-05) < 0.000001
+ assert (features[0][1][-1] - 0.16410418866596271) < 0.000001
+
+
#==============================================================================
def convert_array_to_list_of_frame_cont(data):
"""
@@ -350,6 +419,57 @@ def test_video_svm_pad_algorithm():
assert precision > 0.99
+#==============================================================================
+def test_video_gmm_pad_algorithm():
+ """
+ Test the VideoGmmPadAlgorithm algorithm.
+ """
+
+ random.seed(7)
+
+ N = 1000
+ mu = 1
+ sigma = 1
+ real_array = np.transpose( np.vstack([[random.gauss(mu, sigma) for _ in range(N)], [random.gauss(mu, sigma) for _ in range(N)]]) )
+
+ mu = 5
+ sigma = 1
+ attack_array = np.transpose( np.vstack([[random.gauss(mu, sigma) for _ in range(N)], [random.gauss(mu, sigma) for _ in range(N)]]) )
+
+ real = convert_array_to_list_of_frame_cont(real_array)
+
+ N_COMPONENTS = 1
+ RANDOM_STATE = 3
+ FRAME_LEVEL_SCORES_FLAG = True
+
+ algorithm = VideoGmmPadAlgorithm(n_components = N_COMPONENTS,
+ random_state = RANDOM_STATE,
+ frame_level_scores_flag = FRAME_LEVEL_SCORES_FLAG)
+
+ # training_features[0] - training features for the REAL class.
+ real_array_converted = algorithm.convert_list_of_frame_cont_to_array(real) # output is array
+
+ assert (real_array == real_array_converted).all()
+
+ # Train the GMM machine and get normalizers:
+ machine, features_mean, features_std = algorithm.train_gmm(real = real_array_converted,
+ n_components = algorithm.n_components,
+ random_state = algorithm.random_state)
+
+ algorithm.machine = machine
+
+ algorithm.features_mean = features_mean
+
+ algorithm.features_std = features_std
+
+ scores_real = algorithm.project(real_array_converted)
+
+ scores_attack = algorithm.project(attack_array)
+
+ assert (np.min(scores_real) + 7.9423798970985917) < 0.000001
+ assert (np.max(scores_real) + 1.8380480068281055) < 0.000001
+ assert (np.min(scores_attack) + 38.831260843070098) < 0.000001
+ assert (np.max(scores_attack) + 5.3633030621521272) < 0.000001
diff --git a/bob/pad/face/utils/__init__.py b/bob/pad/face/utils/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/bob/pad/face/utils/face_detection_utils.py b/bob/pad/face/utils/face_detection_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..2214958d80a15c7730517b5a82b49d52ab4a0abe
--- /dev/null
+++ b/bob/pad/face/utils/face_detection_utils.py
@@ -0,0 +1,87 @@
+#!/usr/bin/env python
+# vim: set fileencoding=utf-8 :
+
+"""
+This file contains face detection utils.
+"""
+#==============================================================================
+# Import here:
+
+import bob.ip.dlib # for face detection functionality
+
+
+#==============================================================================
+def detect_face_in_image(image):
+ """
+ This function detects a face in the input image.
+
+ **Parameters:**
+
+ ``image`` : 3D :py:class:`numpy.ndarray`
+ A color image to detect the face in.
+
+ **Returns:**
+
+ ``annotations`` : :py:class:`dict`
+ A dictionary containing annotations of the face bounding box.
+ Dictionary must be as follows ``{'topleft': (row, col), 'bottomright': (row, col)}``.
+ """
+
+ bounding_box, _ = bob.ip.dlib.FaceDetector().detect_single_face(image)
+
+ annotations = {}
+
+ if bounding_box is not None:
+
+ annotations['topleft'] = bounding_box.topleft
+
+ annotations['bottomright'] = bounding_box.bottomright
+
+ else:
+
+ annotations['topleft'] = (0, 0)
+
+ annotations['bottomright'] = (0, 0)
+
+ return annotations
+
+
+#==============================================================================
+def detect_faces_in_video(frame_container):
+ """
+ This function detects a face in each farme of the input video.
+
+ **Parameters:**
+
+ ``frame_container`` : FrameContainer
+ FrameContainer containing the frames data.
+
+ **Returns:**
+
+ ``annotations`` : :py:class:`dict`
+ A dictionary containing the annotations for each frame in the video.
+ Dictionary structure: ``annotations = {'1': frame1_dict, '2': frame1_dict, ...}``.
+ Where ``frameN_dict = {'topleft': (row, col), 'bottomright': (row, col)}``
+ is the dictionary defining the coordinates of the face bounding box in frame N.
+ """
+
+ annotations = {}
+
+ for idx, frame in enumerate(frame_container):
+
+ image = frame[1]
+
+ frame_annotations = detect_face_in_image(image)
+
+ annotations[str(idx)] = frame_annotations
+
+ return annotations
+
+
+
+
+
+
+
+
+
diff --git a/requirements.txt b/requirements.txt
index 4eff3ad6e2749acdbb9af39bbd8e6ca322a2f2e1..6692bce0eb1efa0062578f7755eea27928de0e94 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -12,6 +12,7 @@ bob.bio.video
bob.io.image
bob.ip.color
bob.ip.qualitymeasure
+bob.ip.dlib
bob.learn.libsvm
bob.learn.linear
scikit-learn
diff --git a/setup.py b/setup.py
index 228505a442da3ff6ed99c33d09c2c447d25071ec..0a0578e9d4355fcdc12bd181768fb43c0c237ffe 100644
--- a/setup.py
+++ b/setup.py
@@ -103,6 +103,7 @@ setup(
# registered preprocessors:
'bob.pad.preprocessor': [
'empty-preprocessor = bob.pad.face.config.preprocessor.filename:empty_preprocessor', # no preprocessing
+ 'preprocessor-rgb-face-detect = bob.pad.face.config.preprocessor.video_face_crop:preprocessor_rgb_face_detect', # detect faces locally replacing database annotations
],
# registered extractors: