small fixes

Sports2D/Utilities/skeletons.py

@@ -217,9 +217,17 @@
             ]),
         ]),
     ]),
-    Node("Nose", id=0, children=[
-        Node("REye", id=5),
+    Node("Nose", id=0, children=[ 
+        Node("LEyeInner", id=1),
         Node("LEye", id=2),
+        Node("LEyeOuter", id=3),
+        Node("REyeInner", id=4),
+        Node("REye", id=5),
+        Node("REyeOuter", id=6),
+        Node("LEar", id=7),
+        Node("REar", id=8),
+        Node("LMouth", id=9),
+        Node("RMouth", id=10),
     ]),
     Node("RShoulder", id=12, children=[
         Node("RElbow", id=14, children=[

Sports2D/compute_angles.py

@@ -249,7 +249,7 @@ def segment_angles_series_from_points(df_points, angle_params, segment):
     ang_series = points2D_to_angles(points_list)
     ang_series += angle_params[2]
     ang_series *= angle_params[3]
-    ang_series = np.where(ang_series>180,ang_series-360,ang_series)
+    # ang_series = np.where(ang_series>180,ang_series-360,ang_series)
 
     # For trunk: mean between angles RHip-RShoulder and LHip-LShoulder
     if segment == 'Trunk':
@@ -264,7 +264,10 @@ def segment_angles_series_from_points(df_points, angle_params, segment):
         ang_series += angle_params[2]
         ang_series *= angle_params[3]
         ang_series = np.mean((ang_seriesR, ang_series), axis=0)
-        ang_series = np.where(ang_series>180,ang_series-360,ang_series)
+        # ang_series = np.where(ang_series>180,ang_series-360,ang_series)
+
+    if ang_series.mean() > 180: ang_series -= 360
+    if ang_series.mean() < -180: ang_series += 360
 
     return ang_series
 

Sports2D/detect_pose.py

@@ -213,7 +213,10 @@ def min_with_single_indices(L, T):
     - T_minL: list of tuples associated with smallest values of L
     '''
 
-    minL = [np.min(L)]
+    try:
+        minL = [np.min(L)]
+    except:
+        return [], [], []
     argminL = [np.argmin(L)]
     T_minL = [T[argminL[0]]]
 
@@ -492,9 +495,13 @@ def save_imgvid_reID(video_path, video_result_path, save_vid=1, save_img=1, *pos
 
     # Open csv files
     coords = []
-    for c in csv_paths:
-        with open(c) as c_f:
-            coords += [pd.read_csv(c_f, header=[0,1,2,3])]
+    try:
+        for c in csv_paths:
+            with open(c) as c_f:
+                coords += [pd.read_csv(c_f, header=[0,1,2,3])]
+    except:
+        logging.warning('No csv files found.')
+        return
 
     # Open video frame by frame
     cap = cv2.VideoCapture(str(video_path))
@@ -535,7 +542,8 @@ def save_imgvid_reID(video_path, video_result_path, save_vid=1, save_img=1, *pos
             break
         f += 1
     cap.release()
-    writer.release()
+    if save_vid:
+        writer.release()
 
 
 def detect_pose_fun(config_dict, video_file):
@@ -625,21 +633,29 @@ def detect_pose_fun(config_dict, video_file):
             elif platform == "linux" or platform=="linux2":
                 run_openpose_linux(video_path, json_path, pose_model)
             os.chdir(root_dir)
-
+
+
+    elif pose_algo == 'BLAZEPOSE':
+        pose_model = pose_algo
+        json_path = result_dir / '_'.join((video_file_stem,pose_algo,'json'))
+        model_complexity = config_dict.get('pose').get('BLAZEPOSE').get('model_complexity')
+        Blazepose_runsave.blazepose_detec_func(input_file=video_path, save_images=False, to_json=True, save_video=False, to_csv=False, output_folder=result_dir, model_complexity=model_complexity)
+
     # Sort people and save to csv, optionally display plot
+    try:
         json_to_csv(json_path, frame_rate, pose_model, interp_gap_smaller_than, filter_options, show_plots)
+    except:
+        logging.warning('No person detected or persons could not be associated across frames.')
+        return
 
     # Save images and files after reindentification
-        if save_img and save_vid:
-            logging.info(f'Saving images and video in {result_dir}.')
-        if save_img and not save_vid:
-            logging.info(f'Saving images in {result_dir}.')
-        if not save_img and save_vid:
-            logging.info(f'Saving video in {result_dir}.')
-        if save_vid or save_img:
-            save_imgvid_reID(video_path, video_result_path, save_vid, save_img, pose_model)
+    if save_img and save_vid:
+        logging.info(f'Saving images and video in {result_dir}.')
+    if save_img and not save_vid:
+        logging.info(f'Saving images in {result_dir}.')
+    if not save_img and save_vid:
+        logging.info(f'Saving video in {result_dir}.')
+    if save_vid or save_img:
+        save_imgvid_reID(video_path, video_result_path, save_vid, save_img, pose_model)
 
-
-    elif pose_algo == 'BLAZEPOSE':
-        model_complexity = config_dict.get('pose').get('BLAZEPOSE').get('model_complexity')
-        Blazepose_runsave.blazepose_detec_func(input_file=video_path, save_images=save_img, to_json=True, save_video=save_vid, to_csv=True, output_folder=result_dir, model_complexity=model_complexity)
+

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = sports2d
-version = 0.3.3
+version = 0.3.4
 author = David Pagnon
 author_email = contact@david-pagnon.com
 description = Detect pose and compute 2D joint angles from a video.
@@ -38,7 +38,7 @@ install_requires =
     scipy
     anytree
     opencv-python
-    mediapipe==0.10.0
+    mediapipe
     matplotlib
     PyQt5
     statsmodels