Merge pull request #415 from hdefazio/fix/black

Black formatting

angel_system/activity_classification/tcn_hpl/predict.py

@@ -119,9 +119,9 @@ def normalize_detection_features(
     :param image_height: Integer pixel height of the image that object
         detections were generated on.
     :param num_det_classes: Number of object detection classes (note: DOES include the hand labels but DOES NOT include the patient and user labels)
-    :param normalize_pixel_pts: If true, will apply the NormalizePixelPts data augmentation to 
+    :param normalize_pixel_pts: If true, will apply the NormalizePixelPts data augmentation to
         the ``det_feats``
-    :param normalize_center_pts: If true, will apply the NormalizeFromCenter data augmentation to 
+    :param normalize_center_pts: If true, will apply the NormalizeFromCenter data augmentation to
         the ``det_feats``
 
     :return: Normalized object detection features.
@@ -178,13 +178,13 @@ def objects_to_feats(
     :param pose_memo: Optional memoization cache of the pose used for each
         frame and the repeat pose count at each frame
     :param top_k_objects: Number top confidence objects to use per label, defaults to 1
-    :param normalize_pixel_pts: If true, will apply the NormalizePixelPts data augmentation to 
+    :param normalize_pixel_pts: If true, will apply the NormalizePixelPts data augmentation to
         the feature vector
-    :param normalize_center_pts: If true, will apply the NormalizeFromCenter data augmentation to 
+    :param normalize_center_pts: If true, will apply the NormalizeFromCenter data augmentation to
         the feature vector
     :param pose_repeat_rate: The maximum number of sequential None value poses that can be replaced with
-        a valid pose in a previous frame. If this number is exceeded, the pose 
-        for the frame will remain None. 
+        a valid pose in a previous frame. If this number is exceeded, the pose
+        for the frame will remain None.
 
     :raises ValueError: No object detections nor patient poses passed in.
     :raises ValueError: No non-None object detections in the given input
@@ -227,7 +227,7 @@ def objects_to_feats(
     for i, (pose, detections) in enumerate(
         zip(frame_patient_poses, frame_object_detections)
     ):
-        
+
         if detections is None:
             print("no detections!")
             continue
@@ -268,7 +268,7 @@ def objects_to_feats(
             elif last_pose is not None:
                 repeated_pose_count += 1
                 # Repeat at most {pose_repeat_rate} poses in a row
-                if repeated_pose_count > (pose_repeat_rate/2):
+                if repeated_pose_count > (pose_repeat_rate / 2):
                     last_pose = None
                     print("Resetting pose to None")
                     repeated_pose_count = 0
@@ -283,7 +283,7 @@ def objects_to_feats(
 
             pose_memo[memo_key] = {
                 "last_pose": last_pose,
-                "repeated_pose_count": repeated_pose_count
+                "repeated_pose_count": repeated_pose_count,
             }
 
             # Grab the joint keypoints
@@ -316,7 +316,6 @@ def objects_to_feats(
                 .astype(np.float32)
             )
 
-
             feat_memo[memo_key] = feat
 
         feature_ndim = feat.shape

angel_system/activity_classification/train_activity_classifier.py

@@ -113,7 +113,7 @@ def compute_feats(
     ann_by_image: dict,
     feat_version=1,
     top_k_objects=1,
-    pose_repeat_rate=4
+    pose_repeat_rate=4,
 ) -> Tuple[np.ndarray, np.ndarray]:
     """Compute features from object detections
 

angel_system/data/common/create_custom_learn_video_dataset.py

@@ -65,9 +65,9 @@ def main(args):
             ):
                 temp_df = df[df["# 1: Detection or Track-id"] == str(label)]
                 if temp_df.iloc[0]["10-11+: Repeated Species"] not in label_dict.keys():
-                    label_dict[temp_df.iloc[0]["10-11+: Repeated Species"]] = (
-                        label_counter
-                    )
+                    label_dict[
+                        temp_df.iloc[0]["10-11+: Repeated Species"]
+                    ] = label_counter
                     label_counter += 1
                 min_frame = pd.to_numeric(temp_df["3: Unique Frame Identifier"]).min()
                 max_frame = pd.to_numeric(temp_df["3: Unique Frame Identifier"]).max()

angel_system/data/common/kwcoco_utils.py

@@ -92,7 +92,7 @@ def add_activity_gt_to_kwcoco(topic, task, dset, activity_config_fn):
         # Update the activity gt for each image
         for gid in sorted(image_ids):
             im = dset.imgs[gid]
-            
+
             frame_idx, time = time_from_name(im["file_name"], topic)
 
             if time:
@@ -106,7 +106,7 @@ def add_activity_gt_to_kwcoco(topic, task, dset, activity_config_fn):
                 label = "background"
                 activity_label = label
             else:
-                label = matching_gt.iloc[0]["class_label"] # default to the first gt
+                label = matching_gt.iloc[0]["class_label"]  # default to the first gt
 
                 # Hacky temporary fix
                 # In the medical data, step 1 can cover the same frames as other steps,
@@ -134,7 +134,7 @@ def add_activity_gt_to_kwcoco(topic, task, dset, activity_config_fn):
                         f"Label: {label} is not in the activity labels config, ignoring"
                     )
                     print(f"LABEL: {label}, {type(label)}")
-                    
+
                     label = "background"
                     activity_label = label
                 else:

angel_system/global_step_prediction/global_step_predictor.py

@@ -235,9 +235,9 @@ def initialize_new_recipe_tracker(self, recipe, config_fn=None):
                 },
             )
 
-        tracker_dict["last_granular_step_per_broad_step"] = (
-            self.get_last_granular_step_per_broad_step(broad_steps)
-        )
+        tracker_dict[
+            "last_granular_step_per_broad_step"
+        ] = self.get_last_granular_step_per_broad_step(broad_steps)
         tracker_dict["recipe"] = recipe
 
         tracker_dict["current_broad_step"] = 0
@@ -254,9 +254,9 @@ def initialize_new_recipe_tracker(self, recipe, config_fn=None):
         tracker_dict["broad_step_to_activity_ids"] = [
             self.get_unique(step["activity_ids"]) for step in broad_steps
         ]
-        tracker_dict["granular_step_to_activity_id"] = (
-            self.get_activity_per_granular_step(broad_steps)
-        )
+        tracker_dict[
+            "granular_step_to_activity_id"
+        ] = self.get_activity_per_granular_step(broad_steps)
 
         # Labels
         tracker_dict["broad_step_to_label"] = [step["label"] for step in broad_steps]
@@ -300,9 +300,9 @@ def increment_granular_step(self, tracker_ind):
 
         if current_granular_step < num_granular_steps:
             self.trackers[tracker_ind]["current_granular_step"] += 1
-            self.trackers[tracker_ind]["current_broad_step"] = (
-                self.granular_to_broad_step(tracker, current_granular_step)
-            )
+            self.trackers[tracker_ind][
+                "current_broad_step"
+            ] = self.granular_to_broad_step(tracker, current_granular_step)
         elif current_granular_step == num_granular_steps and tracker["active"] == True:
             self.trackers[tracker_ind]["active"] = False
         else:
@@ -335,9 +335,9 @@ def decrement_granular_step(self, tracker_ind):
 
         if current_granular_step > 0:
             self.trackers[tracker_ind]["current_granular_step"] -= 1
-            self.trackers[tracker_ind]["current_broad_step"] = (
-                self.granular_to_broad_step(tracker, current_granular_step)
-            )
+            self.trackers[tracker_ind][
+                "current_broad_step"
+            ] = self.granular_to_broad_step(tracker, current_granular_step)
         else:
             raise Exception(
                 f"Tried to decrement tracker #{tracker_ind}: "

ros/angel_system_nodes/angel_system_nodes/activity_classification/activity_classifier_tcn.py

@@ -420,9 +420,9 @@ def _thread_populate_from_coco(self, input_coco_path: Path) -> None:
                 # Creates [n_det, n_label] matrix, which we assign to and then
                 # ravel into the message slot.
                 conf_mat = np.zeros((n_dets, len(obj_labels)), dtype=np.float64)
-                conf_mat[np.arange(n_dets), image_annots.get("category_id")] = (
-                    image_annots.get("confidence")
-                )
+                conf_mat[
+                    np.arange(n_dets), image_annots.get("category_id")
+                ] = image_annots.get("confidence")
                 det_msg.label_confidences.extend(conf_mat.ravel())
 
             # Calling the image callback last since image frames define the
@@ -602,7 +602,7 @@ def rt_loop(self):
 
                 window = self._buffer.get_window(
                     self._window_size,
-                    have_leading_object=self._window_lead_with_objects
+                    have_leading_object=self._window_lead_with_objects,
                 )
 
                 # log.info(f"buffer contents: {window.obj_dets}")
@@ -762,11 +762,11 @@ def _process_window(self, window: InputWindow) -> ActivityDetection:
                 feat_version=self._feat_version,
                 image_width=self._img_pix_width,
                 image_height=self._img_pix_height,
-                #feature_memo=memo_object_to_feats, # passed by reference so this gets updated in the function and changes persist here
-                #pose_memo=queued_pose_memo,
+                # feature_memo=memo_object_to_feats, # passed by reference so this gets updated in the function and changes persist here
+                # pose_memo=queued_pose_memo,
                 normalize_pixel_pts=self.model_normalize_pixel_pts,
                 normalize_center_pts=self.model_normalize_center_pts,
-                pose_repeat_rate=self._pose_repeat_rate
+                pose_repeat_rate=self._pose_repeat_rate,
             )
         except ValueError as ex:
             log.warn(f"object-to-feats: ValueError: {ex}")

ros/angel_utils/python/angel_utils/activity_classification.py

@@ -63,14 +63,20 @@ def __repr__(self):
         timestamp frame and correlated data **lower** in the table (higher
         index). This order is arbitrary.
         """
-        return repr(pd.DataFrame(
-            data={
-                "frames": [time_to_int(f[0]) for f in self.frames],
-                "detections": [(d.num_detections if d else None) for d in self.obj_dets],
-                "poses": [(len(p.joints) if p else None) for p in self.patient_joint_kps],
-            },
-            dtype=pd.Int64Dtype,
-        ))
+        return repr(
+            pd.DataFrame(
+                data={
+                    "frames": [time_to_int(f[0]) for f in self.frames],
+                    "detections": [
+                        (d.num_detections if d else None) for d in self.obj_dets
+                    ],
+                    "poses": [
+                        (len(p.joints) if p else None) for p in self.patient_joint_kps
+                    ],
+                },
+                dtype=pd.Int64Dtype,
+            )
+        )
 
 
 # TODO: A more generic version of InputBuffer
@@ -337,11 +343,13 @@ def get_window(
             # starting.
             # Finally, the extracted slice is reversed in order again so that
             # the final list is in temporally ascending order.
-            window_frames = list(itertools.islice(
-                reversed(self.frames),
-                window_frame_start_idx,
-                window_size + window_frame_start_idx
-            ))[::-1]
+            window_frames = list(
+                itertools.islice(
+                    reversed(self.frames),
+                    window_frame_start_idx,
+                    window_size + window_frame_start_idx,
+                )
+            )[::-1]
             window_frame_times: List[Time] = [wf[0] for wf in window_frames]
             window_frame_times_ns: List[int] = [
                 time_to_int(wft) for wft in window_frame_times

ros/angel_utils/scripts/bag_extractor.py

@@ -153,19 +153,19 @@ def __init__(self):
         if self.extract_head_pose_data or self.extract_depth_head_pose_data:
             self.msg_type_to_handler_map[HeadsetPoseData] = self.handle_head_pose_msg
         if self.extract_hand_pose_data:
-            self.msg_type_to_handler_map[HandJointPosesUpdate] = (
-                self.handle_hand_pose_msg
-            )
+            self.msg_type_to_handler_map[
+                HandJointPosesUpdate
+            ] = self.handle_hand_pose_msg
         if self.extract_spatial_map_data:
             self.msg_type_to_handler_map[SpatialMesh] = self.handle_spatial_mesh_msg
         if self.extract_annotation_event_data:
-            self.msg_type_to_handler_map[AnnotationEvent] = (
-                self.handle_annotation_event_msg
-            )
+            self.msg_type_to_handler_map[
+                AnnotationEvent
+            ] = self.handle_annotation_event_msg
         if self.extract_activity_detection_data:
-            self.msg_type_to_handler_map[ActivityDetection] = (
-                self.handle_activity_detection_msg
-            )
+            self.msg_type_to_handler_map[
+                ActivityDetection
+            ] = self.handle_activity_detection_msg
         if self.extract_task_update_data:
             self.msg_type_to_handler_map[TaskUpdate] = self.handle_task_update_msg