Fixed several typos

research/autoaugment/augmentation_transforms.py

@@ -45,7 +45,7 @@ def zero_pad_and_crop(img, amount=4):
 
   Args:
     img: numpy image that will be zero padded and cropped.
-    amount: amount of zeros to pad `img` with horizontally and verically.
+    amount: amount of zeros to pad `img` with horizontally and vertically.
 
   Returns:
     The cropped zero padded img. The returned numpy array will be of the same

research/autoaugment/train_cifar.py

@@ -353,7 +353,7 @@ def _compute_final_accuracies(self, meval):
     return valid_accuracy, test_accuracy
 
   def run_model(self):
-    """Trains and evalutes the image model."""
+    """Trains and evaluates the image model."""
     hparams = self.hparams
 
     # Build the child graph

research/cognitive_planning/envs/active_vision_dataset_env.py

@@ -195,7 +195,7 @@ def read_all_poses(dataset_root, world):
     id of each view and the values are tuple of (x, z, R, scale). Where x and z
     are the first and third coordinate of translation. R is the 3x3 rotation
     matrix and scale is a float scalar that indicates the scale that needs to
-    be multipled to x and z in order to get the real world coordinates.
+    be multiplied to x and z in order to get the real world coordinates.
 
   Raises:
     ValueError: if the number of images do not match the number of poses read.
@@ -675,7 +675,7 @@ def _step_no_reward(self, action):
 
     Returns:
       observation, done, info
-      observation: dictonary that contains all the observations specified in
+      observation: dictionary that contains all the observations specified in
         modality_types.
         observation[task_env.ModalityTypes.OBJECT_DETECTION]: contains the
         detection of the current view.
@@ -897,7 +897,7 @@ def _reset_env(
       self._vertex_to_pose[self._cur_world] = {
           index: (-index,) for index in self._cur_graph.target_indexes.values()
       }
-      # Calling vetex_to_pose for each vertex results in filling out the
+      # Calling vertex_to_pose for each vertex results in filling out the
       # dictionaries that contain pose related data.
       for image_id in self._world_id_dict[self._cur_world]:
         self.vertex_to_pose(self.to_vertex(image_id))
@@ -951,7 +951,7 @@ def vertex_to_pose(self, v):
       v: integer, vertex index.
 
     Returns:
-      (x, z, dir_x, dir_z) where x and z are the tranlation and dir_x, dir_z are
+      (x, z, dir_x, dir_z) where x and z are the translation and dir_x, dir_z are
         a vector giving direction of the view.
     """
     if v in self._vertex_to_pose[self._cur_world]:
@@ -996,7 +996,7 @@ def check_scene_graph(self, world, goal):
     """
     obs = self._reset_env(new_world=world, new_goal=goal)
     if not obs:
-      print '{} is not availble in {}'.format(goal, world)
+      print '{} is not available in {}'.format(goal, world)
       return True
     for image_id in self._world_id_dict[self._cur_world]:
       print 'check image_id = {}'.format(image_id)

research/cognitive_planning/preprocessing/cifarnet_preprocessing.py

@@ -41,7 +41,7 @@ def preprocess_for_train(image,
     image: A `Tensor` representing an image of arbitrary size.
     output_height: The height of the image after preprocessing.
     output_width: The width of the image after preprocessing.
-    padding: The amound of padding before and after each dimension of the image.
+    padding: The amount of padding before and after each dimension of the image.
     add_image_summaries: Enable image summaries.
 
   Returns:

research/cognitive_planning/tasks.py

@@ -80,7 +80,7 @@ def classification_loss(truth, predicted, weights=None, is_one_hot=True):
 
   Computes the mean of cross entropy losses for all pairs of true labels and
   predictions. It wraps around a tf implementation of the cross entropy loss
-  with additional reformating of the inputs. If the truth and predicted are
+  with additional reformatting of the inputs. If the truth and predicted are
   n-rank Tensors with n > 2, then these are reshaped to 2-rank Tensors. It
   allows for truth to be specified as one hot vector or class indices. Finally,
   a weight can be specified for each element in truth and predicted.
@@ -489,7 +489,7 @@ def _obs_to_state(self, path, states):
         [np.reshape(p, [1, 2]) for p in path_coordinates])
 
     # The observations are taken along a smoothed trajectory following the path.
-    # We compute a mapping between the obeservations and the map vertices.
+    # We compute a mapping between the observations and the map vertices.
     path_to_obs = collections.defaultdict(list)
     obs_to_state = []
     for i, s in enumerate(states):
@@ -634,7 +634,7 @@ def episode(self):
     assert len(observations.values()[0]) == len(states)
 
     # The observations are taken along a smoothed trajectory following the path.
-    # We compute a mapping between the obeservations and the map vertices.
+    # We compute a mapping between the observations and the map vertices.
     path_to_obs, obs_to_path = self._obs_to_state(path, states)
 
     # Go over all observations, and sample a query. With probability 0.5 this
@@ -751,7 +751,7 @@ def episode(self):
         2 x observation_size X observation_size x 3 containing a pair of query
         images.
       A tuple of size two. First element is a numpy array of size 2 containing
-        a one hot vector of whether the two observations are neighobring. Second
+        a one hot vector of whether the two observations are neighboring. Second
         element is a boolean numpy value denoting whether this is a valid
         episode.
     """
@@ -1166,7 +1166,7 @@ def _compute_gt_value(self, vertex, target_vertices):
     closer to the goal the value is 1. In case, it takes the agent one step away
     from the goal it would be -1. If it leads to collision or if the agent uses
     action stop before reaching to the goal it is -2. To avoid scale issues the
-    gt_values are multipled by 0.5.
+    gt_values are multiplied by 0.5.
 
     Args:
       vertex: integer, the index of current vertex.
@@ -1355,7 +1355,7 @@ def _compute_gt_value(self, vertex, target_vertices):
     closer to the goal the value is 1. In case, it takes the agent one step away
     from the goal it would be -1. If it leads to collision or if the agent uses
     action stop before reaching to the goal it is -2. To avoid scale issues the
-    gt_values are multipled by 0.5.
+    gt_values are multiplied by 0.5.
 
     Args:
       vertex: integer, the index of current vertex.

research/cognitive_planning/train_supervised_active_vision.py

@@ -88,7 +88,7 @@
 flags.DEFINE_integer('save_summaries_secs', 300,
                      'number of seconds between saving summaries')
 flags.DEFINE_integer('save_interval_secs', 300,
-                     'numer of seconds between saving variables')
+                     'number of seconds between saving variables')
 flags.DEFINE_integer('log_every_n_steps', 20, 'number of steps between logging')
 flags.DEFINE_string('modality_types', '',
                     'modality names in _ separated format')
@@ -122,8 +122,8 @@
 flags.DEFINE_bool('freeze_resnet_weights', True, '')
 flags.DEFINE_string(
     'eval_init_points_file_name', '',
-    'Name of the file that containts the initial locations and'
-    'worlds for each evalution point')
+    'Name of the file that contains the initial locations and'
+    'worlds for each evaluation point')
 
 FLAGS = flags.FLAGS
 TRAIN_WORLDS = [
@@ -377,7 +377,7 @@ def init(sequence_length, eval_init_points_file_name, worlds):
   logging.info('modality types: %r', modality_types)
   # negative reward_goal_range prevents the env from terminating early when the
   # agent is close to the goal. The policy should keep the agent until the end
-  # of the 100 steps either through chosing stop action or oscilating around
+  # of the 100 steps either through choosing stop action or oscillating around
   # the target.
 
   env = active_vision_dataset_env.ActiveVisionDatasetEnv(

research/cognitive_planning/visualization_utils.py

@@ -334,8 +334,8 @@ def draw_bounding_boxes_on_image_tensors(images,
       with keypoints.
     max_boxes_to_draw: Maximum number of boxes to draw on an image. Default 20.
     min_score_thresh: Minimum score threshold for visualization. Default 0.2.
-    use_normalized_coordinates: Whether to assume boxes and kepoints are in
-      normalized coordinates (as opposed to absolute coordiantes).
+    use_normalized_coordinates: Whether to assume boxes and keypoints are in
+      normalized coordinates (as opposed to absolute coordinates).
       Default is True.
 
   Returns:
@@ -402,8 +402,8 @@ def draw_side_by_side_evaluation_image(eval_dict,
     category_index: A category index (dictionary) produced from a labelmap.
     max_boxes_to_draw: The maximum number of boxes to draw for detections.
     min_score_thresh: The minimum score threshold for showing detections.
-    use_normalized_coordinates: Whether to assume boxes and kepoints are in
-      normalized coordinates (as opposed to absolute coordiantes).
+    use_normalized_coordinates: Whether to assume boxes and keypoints are in
+      normalized coordinates (as opposed to absolute coordinates).
       Default is True.
 
   Returns: