Merge pull request #346 from Visual-Behavior/fix_depth_encode_abs

adapt torch1.13

aloscene/depth.py

@@ -46,7 +46,6 @@ def __new__(
         names=("C", "H", "W"),
         **kwargs
     ):
-
         if isinstance(x, str):
             x = load_depth(x)
             names = ("C", "H", "W")
@@ -152,10 +151,10 @@ def encode_absolute(
             depth = torch.clamp(depth, min=prior_clamp_min, max=prior_clamp_max)
 
         if keep_negative and self.is_planar:
-            depth[torch.unsqueeze((depth < 1e-8) & (depth >= 0), dim=0)] = 1e-8
-            depth[torch.unsqueeze((depth >= -1e-8) & (depth < 0), dim=0)] = -1e-8
+            depth[(depth < 1e-8) & (depth >= 0)] = 1e-8
+            depth[(depth >= -1e-8) & (depth < 0)] = -1e-8
         else:
-            depth[torch.unsqueeze(depth < 1e-8, dim=0)] = 1e-8
+            depth[depth < 1e-8] = 1e-8
 
         depth.scale = scale
         depth.shift = shift
@@ -234,7 +233,11 @@ def as_points3d(
         intrinsic = camera_intrinsic if camera_intrinsic is not None else self.cam_intrinsic
         projection = projection if projection is not None else self.projection
         distortion = distortion if distortion is not None else self.distortion
-        assert projection in ["pinhole", "equidistant", "kumler_bauer"], "Only pinhole, equidistant and kumler_bauer are supported."
+        assert projection in [
+            "pinhole",
+            "equidistant",
+            "kumler_bauer",
+        ], "Only pinhole, equidistant and kumler_bauer are supported."
 
         # if self is not planar depth, we must convert to planar depth before projecting to 3d points
         if self.is_planar: