Merge branch 'dev_branch' into 'master'

Added a face detection functionality in the face cropping preprocessor

See merge request !21

bob/pad/face/config/grid.py

@@ -5,22 +5,22 @@ from bob.bio.base.grid import Grid
 
 # Configuration to run on computation cluster:
 idiap = Grid(
-    training_queue='32G',
+    training_queue='8G-io-big',
 
     number_of_preprocessing_jobs=32,
     preprocessing_queue='4G-io-big',
 
     number_of_extraction_jobs=32,
-    extraction_queue='8G-io-big',
+    extraction_queue='4G-io-big',
 
     number_of_projection_jobs=32,
-    projection_queue='8G-io-big',
+    projection_queue='4G-io-big',
 
     number_of_enrollment_jobs=32,
-    enrollment_queue='8G-io-big',
+    enrollment_queue='4G-io-big',
 
     number_of_scoring_jobs=1,
-    scoring_queue='8G-io-big',
+    scoring_queue='4G-io-big',
     )
 
 # Configuration to run on user machines:

bob/pad/face/config/preprocessor/video_face_crop.py

@@ -17,15 +17,50 @@ 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
+FACE_DETECTION_METHOD = "dlib"
+
+rgb_face_detector_dlib = 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,
+                                       face_detection_method = FACE_DETECTION_METHOD)
+
+
+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
+FACE_DETECTION_METHOD = "mtcnn"
+
+rgb_face_detector_mtcnn = 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,
+                                       face_detection_method = FACE_DETECTION_METHOD)
+
+
+
+
+
 
-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)

bob/pad/face/preprocessor/VideoFaceCrop.py

@@ -84,6 +84,12 @@ class VideoFaceCrop(Preprocessor, object):
         cropping.
         Default: ``False``.
 
+    ``face_detection_method`` : :py:class:`str`
+        A package to be used for face detection. Options supported by this
+        package: "dlib" (dlib is a dependency of this package). If  bob.ip.mtcnn
+        is installed in your system you can use it as-well (bob.ip.mtcnn is NOT
+        a dependency of this package).
+
     ``kwargs``
         Remaining keyword parameters passed to the Base constructor, such as ``color_channel`` or ``dtype``.
     """
@@ -101,6 +107,7 @@ class VideoFaceCrop(Preprocessor, object):
                  use_local_cropper_flag = False,
                  rgb_output_flag = False,
                  detect_faces_flag = False,
+                 face_detection_method = "dlib",
                  **kwargs):
 
         super(VideoFaceCrop, self).__init__(cropped_image_size = cropped_image_size,
@@ -114,6 +121,7 @@ class VideoFaceCrop(Preprocessor, object):
                                             use_local_cropper_flag = use_local_cropper_flag,
                                             rgb_output_flag = rgb_output_flag,
                                             detect_faces_flag = detect_faces_flag,
+                                            face_detection_method = face_detection_method,
                                             **kwargs)
 
         self.cropped_image_size = cropped_image_size
@@ -127,6 +135,7 @@ class VideoFaceCrop(Preprocessor, object):
         self.use_local_cropper_flag = use_local_cropper_flag
         self.rgb_output_flag = rgb_output_flag
         self.detect_faces_flag = detect_faces_flag
+        self.face_detection_method = face_detection_method
 
         # Save also the data stored in the kwargs:
         for (k, v) in kwargs.items():
@@ -276,7 +285,7 @@ class VideoFaceCrop(Preprocessor, object):
 
         if self.detect_faces_flag:
 
-            annotations = detect_faces_in_video(frames)
+            annotations = detect_faces_in_video(frames, self.face_detection_method)
 
         if len(frames) != len(annotations): # if some annotations are missing
 
@@ -307,7 +316,9 @@ class VideoFaceCrop(Preprocessor, object):
             name of the file.
         """
 
-        self.video_preprocessor.write_data(frames, file_name)
+        if frames: # save file if FrameContainer is not empty, otherwise do nothing.
+
+            self.video_preprocessor.write_data(frames, file_name)
 
 
     #==========================================================================

bob/pad/face/test/test.py

@@ -167,6 +167,7 @@ def test_video_face_crop():
     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
+    FACE_DETECTION_METHOD = "dlib"
 
     preprocessor = VideoFaceCrop(cropped_image_size = CROPPED_IMAGE_SIZE,
                                  cropped_positions = CROPPED_POSITIONS,
@@ -178,7 +179,8 @@ def test_video_face_crop():
                                  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)
+                                 detect_faces_flag = DETECT_FACES_FLAG,
+                                 face_detection_method = FACE_DETECTION_METHOD)
 
     video, _ = convert_image_to_video_data(image, annotations, 3)
 

bob/pad/face/utils/face_detection_utils.py

@@ -7,11 +7,11 @@ This file contains face detection utils.
 #==============================================================================
 # Import here:
 
-import bob.ip.dlib # for face detection functionality
+import importlib
 
 
 #==============================================================================
-def detect_face_in_image(image):
+def detect_face_in_image(image, method = "dlib"):
     """
     This function detects a face in the input image.
 
@@ -20,34 +20,45 @@ def detect_face_in_image(image):
     ``image`` : 3D :py:class:`numpy.ndarray`
         A color image to detect the face in.
 
+    ``method`` : :py:class:`str`
+        A package to be used for face detection. Options supported by this
+        package: "dlib" (dlib is a dependency of this package). If  bob.ip.mtcnn
+        is installed in your system you can use it as-well (bob.ip.mtcnn is NOT
+        a dependency of this package).
+
     **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)}``.
+        If no annotations found an empty dictionary is returned.
     """
 
-    bounding_box, _ = bob.ip.dlib.FaceDetector().detect_single_face(image)
+    try:
+        face_detection_module = importlib.import_module("bob.ip." + method)
+    except ImportError:
+        raise ImportError("No module named bob.ip." + method)
 
-    annotations = {}
+    if not hasattr(face_detection_module, 'FaceDetector'):
+        raise AttributeError("bob.ip." + method + " module has no attribute FaceDetector")
 
-    if bounding_box is not None:
+    data = face_detection_module.FaceDetector().detect_single_face(image)
 
-        annotations['topleft'] = bounding_box.topleft
+    annotations = {}
 
-        annotations['bottomright'] = bounding_box.bottomright
+    if ( data is not None ) and ( not all([x is None for x in data]) ):
 
-    else:
+        bounding_box = data[0]
 
-        annotations['topleft'] = (0, 0)
+        annotations['topleft'] = bounding_box.topleft
 
-        annotations['bottomright'] = (0, 0)
+        annotations['bottomright'] = bounding_box.bottomright
 
     return annotations
 
 
 #==============================================================================
-def detect_faces_in_video(frame_container):
+def detect_faces_in_video(frame_container, method = "dlib"):
     """
     This function detects a face in each farme of the input video.
 
@@ -56,6 +67,12 @@ def detect_faces_in_video(frame_container):
     ``frame_container`` : FrameContainer
         FrameContainer containing the frames data.
 
+    ``method`` : :py:class:`str`
+        A package to be used for face detection. Options supported by this
+        package: "dlib" (dlib is a dependency of this package). If  bob.ip.mtcnn
+        is installed in your system you can use it as-well (bob.ip.mtcnn is NOT
+        a dependency of this package).
+
     **Returns:**
 
     ``annotations`` : :py:class:`dict`
@@ -63,6 +80,7 @@ def detect_faces_in_video(frame_container):
         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.
+        If no annotations found an empty dictionary is returned.
     """
 
     annotations = {}
@@ -71,9 +89,11 @@ def detect_faces_in_video(frame_container):
 
         image = frame[1]
 
-        frame_annotations = detect_face_in_image(image)
+        frame_annotations = detect_face_in_image(image, method)
+
+        if frame_annotations:
 
-        annotations[str(idx)] = frame_annotations
+            annotations[str(idx)] = frame_annotations
 
     return annotations
 

develop.cfg

@@ -22,6 +22,7 @@ eggs = bob.pad.face
        bob.learn.activation
        bob.ip.flandmark
        bob.ip.facedetect
+       bob.ip.dlib
        bob.ip.qualitymeasure
        bob.learn.linear
        bob.db.base
@@ -56,6 +57,7 @@ develop = src/bob.extension
           src/bob.learn.activation
           src/bob.ip.flandmark
           src/bob.ip.facedetect
+          src/bob.ip.dlib
           src/bob.ip.qualitymeasure
           src/bob.learn.linear
           src/bob.db.base
@@ -92,6 +94,7 @@ bob.io.matlab = git git@gitlab.idiap.ch:bob/bob.io.matlab
 bob.learn.activation = git git@gitlab.idiap.ch:bob/bob.learn.activation
 bob.ip.flandmark = git git@gitlab.idiap.ch:bob/bob.ip.flandmark
 bob.ip.facedetect = git git@gitlab.idiap.ch:bob/bob.ip.facedetect
+bob.ip.dlib = git git@gitlab.idiap.ch:bob/bob.ip.dlib
 bob.ip.qualitymeasure = git git@gitlab.idiap.ch:bob/bob.ip.qualitymeasure
 bob.learn.linear = git git@gitlab.idiap.ch:bob/bob.learn.linear
 bob.db.base = git git@gitlab.idiap.ch:bob/bob.db.base

setup.py

@@ -103,7 +103,8 @@ 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
+            'rgb-face-detect-dlib = bob.pad.face.config.preprocessor.video_face_crop:rgb_face_detector_dlib', # detect faces locally replacing database annotations
+            'rgb-face-detect-mtcnn = bob.pad.face.config.preprocessor.video_face_crop:rgb_face_detector_mtcnn', # detect faces locally replacing database annotations
             ],
 
         # registered extractors: