Some mixed updates.

scripts/test_skelda_dataset.py

@@ -8,8 +8,9 @@ import matplotlib
 import numpy as np
 import tqdm
 
+# import utils_2d_pose
+import utils_2d_pose_ort as utils_2d_pose
 import test_triangulate
-import utils_2d_pose
 from skelda import evals
 
 sys.path.append("/RapidPoseTriangulation/swig/")

scripts/test_triangulate.py

@@ -9,7 +9,8 @@ import cv2
 import matplotlib
 import numpy as np
 
-import utils_2d_pose
+# import utils_2d_pose
+import utils_2d_pose_ort as utils_2d_pose
 from skelda import utils_pose, utils_view
 
 sys.path.append("/RapidPoseTriangulation/swig/")

scripts/utils_2d_pose_ort.py

@@ -97,9 +97,7 @@ class BaseModel(ABC):
                     ishape = list(self.input_shapes[i])
                     if "batch_size" in ishape:
                         max_batch_size = 10
-                        ishape[0] = np.random.choice(
-                            list(range(1, max_batch_size + 1))
-                        )
+                        ishape[0] = np.random.choice(list(range(1, max_batch_size + 1)))
                     tensor = np.random.random(ishape)
                     tensor = tensor * 255
                 else:
@@ -498,17 +496,12 @@ def get_2d_pose(model, imgs, num_joints=17):
     new_poses = []
     for i in range(len(imgs)):
         img = imgs[i]
-
-        poses = []
         dets = model.predict(img)
-        for pose in dets:
-            pose = np.asarray(pose)
-            poses.append(pose)
 
-        if len(poses) == 0:
-            poses.append(np.zeros([num_joints, 3]))
-
-        poses = np.array(poses)
+        if len(dets) == 0:
+            poses = np.zeros([1, num_joints, 3], dtype=float)
+        else:
+            poses = np.asarray(dets, dtype=float)
         new_poses.append(poses)
 
     return new_poses