Port functional utility fixes from Xiaohang workspace.

prima/utils/evaluate_metric.py

@@ -110,6 +110,8 @@ def compute_pck(self, output: Dict, batch: Dict, pck_threshold: Union[List, None
         if pck_threshold is not None:
             for i in range(len(pck_threshold)):
                 self.pck_threshold_list.append(torch.tensor([pck_threshold[i]] * len(pred_keypoints_2d), dtype=torch.float32))
+        if len(self.pck_threshold_list) == 0:
+            return torch.tensor([], dtype=torch.float32)
 
         pcks = []
         # Use mask area if available, otherwise use full image area
@@ -118,7 +120,7 @@ def compute_pck(self, output: Dict, batch: Dict, pck_threshold: Union[List, None
         else:
             # Use full image area as fallback
             seg_area = torch.tensor([self.image_size * self.image_size] * len(pred_keypoints_2d), dtype=torch.float32).unsqueeze(-1)
-        total_visible = torch.sum(conf, dim=-1)
+        total_visible = torch.sum(conf, dim=-1).clamp_min(1e-6)
         for th in self.pck_threshold_list:
             dist = torch.norm(pred_keypoints_2d - gt_keypoints_2d, dim=-1)
 
@@ -147,7 +149,7 @@ def eval_3d(self, output: Dict, batch: Dict):
         Returns: evaluate metric
         """
         if batch['has_smal_params']["betas"].sum() == 0:
-            return 0., 0., 0., [0., 0.], 0.
+            return 0., 0.
 
         pred_keypoints_3d = output["pred_keypoints_3d"].detach()
         pred_keypoints_3d = pred_keypoints_3d[:, None, :, :]

prima/utils/mesh_renderer.py

@@ -231,7 +231,8 @@ def __call__(self, vertices, camera_translation, image, focal_length, text=None,
             alphaMode='OPAQUE',
             baseColorFactor=baseColorFactor)
 
-        camera_translation[0] *= -1.
+        camera_translation_local = camera_translation.copy()
+        camera_translation_local[0] *= -1.
 
         mesh = trimesh.Trimesh(vertices.copy(), self.faces.copy())
         if side_view:
@@ -248,7 +249,7 @@ def __call__(self, vertices, camera_translation, image, focal_length, text=None,
         scene.add(mesh, 'mesh')
 
         camera_pose = np.eye(4)
-        camera_pose[:3, 3] = camera_translation
+        camera_pose[:3, 3] = camera_translation_local
         camera_center = [image.shape[1] / 2., image.shape[0] / 2.]
         camera = pyrender.IntrinsicsCamera(fx=focal_length, fy=focal_length,
                                            cx=camera_center[0], cy=camera_center[1],

prima/utils/misc.py

@@ -29,13 +29,12 @@ def task_wrapper(task_func: Callable) -> Callable:
     """
 
     def wrap(cfg: DictConfig):
-
-        # apply extra utilities
-        extras(cfg)
-
-        # execute the task
+        start_time = time.time()
         try:
-            start_time = time.time()
+            # apply extra utilities
+            extras(cfg)
+
+            # execute the task
             ret = task_func(cfg=cfg)
         except Exception as ex:
             log.exception("")  # save exception to `.log` file

prima/utils/renderer.py

@@ -202,7 +202,8 @@ def __call__(self,
             alphaMode='OPAQUE',
             baseColorFactor=(*mesh_base_color, 1.0))
 
-        camera_translation[0] *= -1.
+        camera_translation_local = camera_translation.copy()
+        camera_translation_local[0] *= -1.
 
         mesh = trimesh.Trimesh(vertices.copy(), self.faces.copy())
         if side_view:
@@ -219,7 +220,7 @@ def __call__(self,
         scene.add(mesh, 'mesh')
 
         camera_pose = np.eye(4)
-        camera_pose[:3, 3] = camera_translation
+        camera_pose[:3, 3] = camera_translation_local
         camera_center = [image.shape[1] / 2., image.shape[0] / 2.]
         camera = pyrender.IntrinsicsCamera(fx=focal_length_to_use, fy=focal_length_to_use,
                                            cx=camera_center[0], cy=camera_center[1], zfar=1e12)