Fix: tensor vs array type correctness

Summary:
For fg-masking depth, we assumed np.array but passed a Tensor; for defining the default depth_mask, vice versa.

Note that we change the intended behaviour for the latter, assuming that 0s are areas with empty depth. When loading depth masks, we replace NaNs with zeros, so it is sensible. It is not a BC change as that branch would crash if executed. Since there was no reports, I assume no one cared.

Reviewed By: bottler

Differential Revision: D47403588

fbshipit-source-id: 1094104176d7d767a5657b5bbc9f5a0cc9da0ede

Author: shapovalov

pytorch3d/implicitron/dataset/frame_data.py

@@ -583,9 +583,9 @@ def build(
             else None,
         )
 
+        fg_mask_np: Optional[np.ndarray] = None
         mask_annotation = frame_annotation.mask
         if mask_annotation is not None:
-            fg_mask_np: Optional[np.ndarray] = None
             if load_blobs and self.load_masks:
                 fg_mask_np, mask_path = self._load_fg_probability(frame_annotation)
                 frame_data.mask_path = mask_path
@@ -627,7 +627,7 @@ def build(
                 frame_data.depth_map,
                 frame_data.depth_path,
                 frame_data.depth_mask,
-            ) = self._load_mask_depth(frame_annotation, frame_data.fg_probability)
+            ) = self._load_mask_depth(frame_annotation, fg_mask_np)
 
         if load_blobs and self.load_point_clouds and point_cloud is not None:
             pcl_path = self._fix_point_cloud_path(point_cloud.path)
@@ -683,7 +683,7 @@ def _postprocess_image(
     def _load_mask_depth(
         self,
         entry: types.FrameAnnotation,
-        fg_probability: Optional[torch.Tensor],
+        fg_mask: Optional[np.ndarray],
     ) -> Tuple[torch.Tensor, str, torch.Tensor]:
         entry_depth = entry.depth
         dataset_root = self.dataset_root
@@ -693,15 +693,15 @@ def _load_mask_depth(
         depth_map = load_depth(self._local_path(path), entry_depth.scale_adjustment)
 
         if self.mask_depths:
-            assert fg_probability is not None
-            depth_map *= fg_probability
+            assert fg_mask is not None
+            depth_map *= fg_mask
 
         mask_path = entry_depth.mask_path
         if self.load_depth_masks and mask_path is not None:
             mask_path = os.path.join(dataset_root, mask_path)
             depth_mask = load_depth_mask(self._local_path(mask_path))
         else:
-            depth_mask = torch.ones_like(depth_map)
+            depth_mask = (depth_map > 0.0).astype(np.float32)
 
         return torch.tensor(depth_map), path, torch.tensor(depth_mask)