Fix look_at_view_transform when object location is not (0,0,0) (#230)

Summary:
The look_at_view_transform did not give the correct results when the object location `at` was not (0,0,0).

The problem was on computing the cameras' location in world's coordinate `C`. It only took into account the camera position from spherical angles, but ignored the object location in the world's coordinate system. I simply modified the C tensor to take into account the object's location which is not necessarily in the origin.

I ran unit tests and all but 4 failed with the same error message: `RuntimeError: CUDA error: invalid device ordinal`. However the same happens before this patch, so I believe these errors are unrelated.

Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/230

Reviewed By: gkioxari

Differential Revision: D23278126

Pulled By: nikhilaravi

fbshipit-source-id: c06e891bc46de8222325ee7b37aa43cde44648e8

pytorch3d/renderer/cameras.py

@@ -1280,8 +1280,11 @@ def look_at_view_transform(
             dist, elev, azim, at, up, device=device
         )
         dist, elev, azim, at, up = broadcasted_args
-        C = camera_position_from_spherical_angles(
+        C = (
-            dist, elev, azim, degrees=degrees, device=device
+            camera_position_from_spherical_angles(
+                dist, elev, azim, degrees=degrees, device=device
+            )
+            + at
         )
 
     R = look_at_rotation(C, at, up, device=device)

tests/test_cameras.py

@@ -167,6 +167,21 @@ class TestCameraHelpers(TestCaseMixin, unittest.TestCase):
         self.assertTrue(torch.allclose(R, R_default, atol=2e-7))
         self.assertTrue(torch.allclose(t, t_default, atol=2e-7))
 
+    def test_look_at_view_transform_non_default_at_position(self):
+        dist = 1.0
+        elev = 0.0
+        azim = 0.0
+        at = ((1, 1, 1),)
+        # Using passed values for dist, elev, azim, at
+        R, t = look_at_view_transform(dist, elev, azim, at=at)
+        # Using default dist=1.0, elev=0.0, azim=0.0
+        R_default, t_default = look_at_view_transform()
+        # test default = passed = expected
+        # R must be the same, t must be translated by (1,-1,1) with respect to t_default
+        t_trans = torch.tensor([1, -1, 1], dtype=torch.float32).view(1, 3)
+        self.assertTrue(torch.allclose(R, R_default, atol=2e-7))
+        self.assertTrue(torch.allclose(t, t_default + t_trans, atol=2e-7))
+
     def test_camera_position_from_angles_python_scalar(self):
         dist = 2.7
         elev = 90.0