Fixed line lengths

.gitignore

@@ -8,6 +8,7 @@
 *.txt
 *.png
 *.jpg
+*.pgf
 *.xml
 *.avi
 *.mp4

examples/visual_voice_activity_detection/live_demo.py

@@ -3,7 +3,8 @@
 from paz.backend.camera import VideoPlayer, Camera
 import paz.pipelines.detection as dt
 
-parser = argparse.ArgumentParser(description='Visual Voice Activity Detection Live Demonstration')
+parser = argparse.ArgumentParser(description='Visual Voice Activity Detection '
+                                             'Live Demonstration')
 parser.add_argument('-c', '--camera_id', type=int, default=0,
                     help='Camera device ID')
 args = parser.parse_args()

examples/visual_voice_activity_detection/recorded_demo.py

@@ -3,11 +3,16 @@
 from paz.backend.camera import VideoPlayer
 import paz.pipelines.detection as dt
 
-parser = argparse.ArgumentParser(description='Visual Voice Activity Detection Recorded Demonstration')
-parser.add_argument('-i', '--input_path', type=str, default="./demo_video.mp4",
-                    help='Path to the video file to be used as input for the VVAD Pipeline.')
-parser.add_argument('-o', '--output_path', type=str, default="./demo_video_labeled.avi",
-                    help='Path to the video file to be used as output for the VVAD Pipeline.')
+parser = argparse.ArgumentParser(description='Visual Voice Activity Detection '
+                                             'Recorded Demonstration')
+parser.add_argument('-i', '--input_path', type=str,
+                    default="./demo_video.mp4",
+                    help='Path to the video file to be used as input for the '
+                         'VVAD Pipeline.')
+parser.add_argument('-o', '--output_path', type=str,
+                    default="./demo_video_labeled.avi",
+                    help='Path to the video file to be used as output for the '
+                         'VVAD Pipeline.')
 args = parser.parse_args()
 
 pipeline = dt.DetectVVAD()

examples/visual_voice_activity_detection/vvad_lrs3_dataset.py

@@ -16,41 +16,52 @@ class VvadLrs3Dataset(Generator):
     # Arguments
         path: String. Full path to vvadlrs3_faceImages_small.h5 file.
         split: String. Valid option contain 'train', 'validation' or 'test'.
-        validation_split: Float. Percentage of the dataset to be used for validation (valid options between 0.0 to 1.0). Set
-            to 0.0 to disable.
-        test_split: Float. Percentage of the dataset to be used for testing (valid options between 0.0 to 1.0). Set
-            to 0.0 to disable.
-        testing: Boolean. If True, a smaller dataset of 200 samples is used for testing. This is useful for testing
-        evaluating: Boolean. If True, the dataset is used for evaluation. This means that the dataset is not shuffled
-            and the indexes will be stored.
-        reduction_method: String. Valid options are 'cut' or 'reduce'. If 'cut' is selected, the video is cut to the
-            reduction_length. If 'reduce' is selected, reduction_length many single frames of the video is removed form
-            the clip.
-        reduced_length: Float. The length of the video after the reduction_method is applied. Choose None if you
+        validation_split: Float. Percentage of the dataset to be used for
+            validation (valid options between 0.0 to 1.0).
+            Set to 0.0 to disable.
+        test_split: Float. Percentage of the dataset to be used for testing
+            (valid options between 0.0 to 1.0).
+            Set to 0.0 to disable.
+        testing: Boolean. If True, a smaller dataset of 200 samples is used
+            for testing. This is useful for testing
+        evaluating: Boolean. If True, the dataset is used for evaluation.
+            This means that the dataset is not shuffled and the indexes will
+            be stored.
+        reduction_method: String. Valid options are 'cut' or 'reduce'.
+            If 'cut' is selected, the video is cut to the reduction_length.
+            If 'reduce' is selected, reduction_length many single frames of
+            the video is removed form the clip.
+        reduced_length: Float. The length of the video after
+            the reduction_method is applied. Choose None if you
             want to keep the original size. None is the default
 
     # References
         -[VVAD-LRS3](https://www.kaggle.com/datasets/adrianlubitz/vvadlrs3)
     """
     def __init__(
-            self, path='.keras/paz/datasets', split='train', validation_split=0.2, test_split=0.1, testing=False,
-            evaluating=False, reduction_method: Reduction_Method = 'cut', reduced_length=None):
+            self, path='.keras/paz/datasets', split='train',
+            validation_split=0.2, test_split=0.1, testing=False,
+            evaluating=False, reduction_method: Reduction_Method = 'cut',
+            reduced_length=None):
         if split != 'train' and split != 'validation' and split != 'test':
             raise ValueError('Invalid split name')
         if validation_split < 0.0 or validation_split > 1.0:
             raise ValueError('Invalid validation split')
         if test_split < 0.0 or test_split > 1.0:
             raise ValueError('Invalid test split')
         if validation_split + test_split > 1.0:
-            raise ValueError('The sum of val_split and test_split must be less than 1.0')
+            raise ValueError('The sum of val_split and test_split must be '
+                             'less than 1.0')
         options = get_args(Reduction_Method)
-        assert reduction_method in options, f"'{reduction_method}' is not in {options}"
+        assert reduction_method in options, \
+            f"'{reduction_method}' is not in {options}"
 
         path = os.path.join(path, 'vvadlrs3_faceImages_small.h5')
 
         class_names = get_class_names('VVAD_LRS3')
 
-        super(VvadLrs3Dataset, self).__init__(path, split, class_names, 'VVAD_LRS3')
+        super(VvadLrs3Dataset, self).__init__(path, split, class_names,
+                                              'VVAD_LRS3')
         self.validation_split = validation_split
         self.test_split = test_split
         self.use_test_data = testing
@@ -60,10 +71,12 @@ def __init__(
         self.reduced_length = reduced_length
 
         # Get total dataset size and length of the video clips
-        self.total_dataset_size, self.video_length = self.get_data_and_video_size()
+        self.total_dataset_size, self.video_length \
+            = self.get_data_and_video_size()
 
         # Split dataset indexes
-        self.indexes_train, self.indexes_validation, self.indexes_test = self.split_data_indices()
+        self.indexes_train, self.indexes_validation, self.indexes_test \
+            = self.split_data_indices()
 
         # Reduction init
         self.dropout_ids = []
@@ -75,7 +88,7 @@ def __call__(self):
         indexes = []
         if self.split == 'train':
             indexes = self.indexes_train
-            random.shuffle(indexes)  # Use always the same seed so every model gets the same shuffle
+            random.shuffle(indexes)
         elif self.split == 'validation':
             indexes = self.indexes_validation
         elif self.split == 'test':
@@ -95,7 +108,8 @@ def __call__(self):
         data.close()
 
     def get_data_and_video_size(self):
-        """Gets the total size of the dataset and the length of the video clips."""
+        """Gets the total size of the dataset and the length of the video clips.
+        """
         data = h5py.File(self.path, mode='r')
 
         total_size = 0
@@ -111,7 +125,8 @@ def get_data_and_video_size(self):
     def split_data_indices(self):
         """Splits the data indices into train, validation and test."""
         indexes_positive = list(range(self.total_dataset_size // 2))
-        indexes_negative = list(range(self.total_dataset_size // 2, self.total_dataset_size))
+        indexes_negative = list(range(self.total_dataset_size // 2,
+                                      self.total_dataset_size))
 
         random.Random(445363).shuffle(indexes_positive)
         random.Random(848641).shuffle(indexes_negative)
@@ -121,10 +136,12 @@ def split_data_indices(self):
 
         if self.validation_split > 0.0:
             indexes_validation, indexes_positive, indexes_negative = \
-                self.split_by_ratio(self.validation_split, indexes_positive, indexes_negative)
+                self.split_by_ratio(self.validation_split, indexes_positive,
+                                    indexes_negative)
         if self.test_split > 0.0:
             indexes_test, indexes_positive, indexes_negative = \
-                self.split_by_ratio(self.test_split, indexes_positive, indexes_negative)
+                self.split_by_ratio(self.test_split, indexes_positive,
+                                    indexes_negative)
 
         indexes_train = indexes_positive + indexes_negative
 
@@ -134,12 +151,14 @@ def split_by_ratio(self, split_ratio, indexes_positive, indexes_negative):
         """Splits the indexes positive and negative by a ratio.
 
         # Arguments
-            split_ratio: Float. Ratio of the split. Valid options are between 0.0 and 1.0.
+            split_ratio: Float. Ratio of the split. Valid options are
+                between 0.0 and 1.0.
             indexes_positive: List of integers. Indexes of the positive samples.
             indexes_negative: List of integers. Indexes of the negative samples.
         """
         split_size = int(split_ratio * 0.5 * self.total_dataset_size)
-        indexes_split = indexes_positive[:split_size] + indexes_negative[:split_size]
+        indexes_split = (indexes_positive[:split_size] +
+                         indexes_negative[:split_size])
         indexes_positive = indexes_positive[split_size:]
         indexes_negative = indexes_negative[split_size:]
         return indexes_split, indexes_positive, indexes_negative
@@ -148,18 +167,22 @@ def reduction_init(self):
         """Initializes the reduction method.
 
         # Arguments
-            reduction_method: String. Valid options are 'cut' or 'reduce'. If 'cut' is selected, the video is cut to the
-                reduction_length. If 'reduce' is selected, reduction_length many single frames of the video is removed
-                form the clip.
-            reduced_length: Float. The length of the video after the reduction_method is applied. Choose None if you
-                want to keep the original size. None is the default
+            reduction_method: String. Valid options are 'cut' or 'reduce'.
+                If 'cut' is selected, the video is cut to the reduction_length.
+                If 'reduce' is selected, reduction_length many single frames of
+                the video is removed form the clip.
+            reduced_length: Float. The length of the video after
+            the reduction_method is applied. Choose None if you want to keep
+            the original size. None is the default.
         """
         if self.reduced_length is None:
             self.reduced_length = self.video_length
         elif self.reduced_length > self.video_length:
-            raise ValueError('reduction_length must be smaller than the length of the video')
+            raise ValueError('reduction_length must be smaller than the length '
+                             'of the video')
         else:
-            self.reduced_length = math.ceil(self.reduced_length / 25 * self.video_length)
+            self.reduced_length = math.ceil(self.reduced_length / 25 *
+                                            self.video_length)
 
         if 'reduce' in self.reduction_method:
             if self.reduced_length == self.video_length:
@@ -169,7 +192,8 @@ def reduction_init(self):
 
                 cal_drop_every = self.video_length / count_dropouts
 
-                self.dropout_ids = [int(i * cal_drop_every - (cal_drop_every / 2))
+                self.dropout_ids = [int(i * cal_drop_every -
+                                        (cal_drop_every / 2))
                                     for i in range(1, count_dropouts + 1)]
 
     def generate_x_data(self, x_train, i):
@@ -183,27 +207,32 @@ def generate_x_data(self, x_train, i):
         if self.evaluating:
             self.index.append(i)
 
-        if 'reduce' in self.reduction_method:  # First tested using appending all wanted frames, but it is
-            # more efficient to remove the unwanted frames
+        if 'reduce' in self.reduction_method:
             x_out = x_train[i]
             x_out = np.delete(x_out, self.dropout_ids, 0)
         else:
             x_out = x_train[i][:self.reduced_length]
         return x_out
 
     def get_index(self):
-        """Gets the index of the current sample. Only available if evaluating is set to True."""
+        """Gets the index of the current sample. Only available if evaluating
+        is set to True.
+        """
         return self.index.pop(0)
 
     def __len__(self):
         if self.total_dataset_size == -1:
-            raise ValueError('You need to call __call__ first to set the total_size')
+            raise ValueError('You need to call __call__ first to set the '
+                             'total_size')
 
         if self.split == 'train':
-            return (self.total_dataset_size - int(self.validation_split * 0.5 * self.total_dataset_size) * 2 -
+            return (self.total_dataset_size -
+                    int(self.validation_split * 0.5 *
+                        self.total_dataset_size) * 2 -
                     int(self.test_split * 0.5 * self.total_dataset_size) * 2)
         elif self.split == 'validation':
-            return int(self.validation_split * 0.5 * self.total_dataset_size) * 2
+            return int(self.validation_split * 0.5 *
+                       self.total_dataset_size) * 2
         elif self.split == 'test':
             print('total_size_test', self.total_dataset_size)
             return int(self.test_split * 0.5 * self.total_dataset_size) * 2