Merge branch 'vito'

sumitkumarjethani · Aug 23, 2023 · b4891bc · b4891bc
2 parents 0e183c6 + 3b191b1
commit b4891bc
Show file tree

Hide file tree

Showing 4 changed files with 25 additions and 25 deletions.
diff --git a/scripts/dataset_from_videos.py b/scripts/dataset_from_videos.py
@@ -44,7 +44,7 @@ def video_into_images(input, output, frames_freq=1, max_frames=10):
 
         if n_frames % frames_freq == 0:
             output_name = output + f"_frame_{n_frames}.jpg"
-            save_image(output_name, frame)
+            save_image(frame, output_name)
         n_frames += 1
 
 

diff --git a/src/fall_detection/fall/data.py b/src/fall_detection/fall/data.py
@@ -56,13 +56,12 @@ def load_pose_samples_from_dir(
         with open(os.path.join(landmarks_dir, file_name)) as csv_file:
             csv_reader = csv.reader(csv_file, delimiter=file_separator)
             for row in csv_reader:
-                if len(row) == 0: continue
+                if len(row) == 0:
+                    continue
                 assert (
                     len(row) == n_landmarks * 3 + 1
                 ), "Wrong number of values: {}".format(len(row))
-                landmarks = np.array(row[1:], np.float32).reshape(
-                    [n_landmarks, 3]
-                )
+                landmarks = np.array(row[1:], np.float32).reshape([n_landmarks, 3])
                 pose_samples.append(
                     PoseSample(
                         name=row[0],

diff --git a/src/fall_detection/pose/data.py b/src/fall_detection/pose/data.py
@@ -18,7 +18,7 @@ def __init__(
         images_in_folder,
         images_out_folder,
         csvs_out_folder,
-        pose_augmentators: List[PoseAugmentation],
+        pose_augmentators: List[PoseAugmentation] = [],
         per_pose_class_limit=None,
     ):
         self._images_in_folder = images_in_folder
@@ -81,42 +81,43 @@ def __call__(self, pose_model: PoseModel):
                 # Bootstrap every image.
                 for image_name in tqdm.tqdm(image_names):
                     # Load image.
-                    initial_frame = cv2.imread(os.path.join(images_in_folder, image_name))
+                    initial_frame = cv2.imread(
+                        os.path.join(images_in_folder, image_name)
+                    )
 
                     base_image_name = ".".join(image_name.split(".")[:-1])
-                    image_extension = image_name.split('.')[-1]
+                    image_extension = image_name.split(".")[-1]
 
                     # Input frames
                     input_frames = [(image_name, initial_frame)]
-                    
+
                     # Check if any pose augmentations required
-                    if len(self._pose_augmentators):
+                    if len(self._pose_augmentators) > 0:
                         for pose_augmentator in self._pose_augmentators:
-                            input_frames.append((
-                                f"{base_image_name}_{pose_augmentator.get_pose_augmentaion_name()}.{image_extension}",
-                                pose_augmentator(initial_frame)
-                            ))
-
+                            input_frames.append(
+                                (
+                                    f"{base_image_name}_{pose_augmentator.get_pose_augmentaion_name()}.{image_extension}",
+                                    pose_augmentator(initial_frame),
+                                )
+                            )
+
                     for input_frame_name, input_frame in input_frames:
                         # Initialize fresh pose tracker and run it.
                         results = pose_model.predict(input_frame)
 
                         # Save image with pose prediction (if pose was detected).
                         output_frame = input_frame.copy()
 
-<<<<<<< HEAD
-                    cv2.imwrite(
-                        os.path.join(images_out_folder, image_name), output_frame
-                    )
-=======
                         if results is not None:
                             output_frame = pose_model.draw_landmarks(
                                 image=output_frame,
                                 results=results,
                             )
->>>>>>> b1c4cb19cf767b2b66c29dad46b92007d58821c4
 
-                        cv2.imwrite(os.path.join(images_out_folder, input_frame_name), output_frame)
+                        cv2.imwrite(
+                            os.path.join(images_out_folder, input_frame_name),
+                            output_frame,
+                        )
 
                         # Save landmarks if pose was detected.
                         if results is not None:
@@ -129,7 +130,8 @@ def __call__(self, pose_model: PoseModel):
                                 results, frame_height, frame_width
                             )
                             csv_out_writer.writerow(
-                                [input_frame_name] + pose_landmarks.flatten().astype(str).tolist()
+                                [input_frame_name]
+                                + pose_landmarks.flatten().astype(str).tolist()
                             )
         self.align_images_and_csvs()
 

diff --git a/tests/test_fall.py b/tests/test_fall.py
@@ -199,7 +199,7 @@ def test_fall_pipeline():
 
     pose_sample_generator(pose_model=pose_model)
 
-    embedder = PoseEmbedder(landmark_names=COCO_POSE_KEYPOINTS)
+    embedder = PoseEmbedder(landmark_names=COCO_POSE_KEYPOINTS, dims=2)
 
     classifier = EstimatorClassifier(
         estimator=make_pipeline(StandardScaler(), LogisticRegression(random_state=42)),
@@ -208,7 +208,6 @@ def test_fall_pipeline():
 
     pose_samples = load_pose_samples_from_dir(
         pose_embedder=embedder,
-        n_dimensions=3,
         n_landmarks=17,
         landmarks_dir="./tests/test_data/test_dataset_csv",
         file_extension="csv",