Efficient PnP weighting bug fix

Summary:
There is a bug in efficient PnP that incorrectly weights points. This fixes it.

The test does not pass for the previous version with the bug.

Reviewed By: shapovalov

Differential Revision: D22449357

fbshipit-source-id: f5a22081e91d25681a6a783cce2f5c6be429ca6a

pytorch3d/ops/perspective_n_points.py

@@ -66,6 +66,10 @@ def _build_M(y, alphas, weight):
     def prepad(t, v):
         return F.pad(t, (1, 0), value=v)
 
+    if weight is not None:
+        # weight the alphas in order to get a correctly weighted version of M
+        alphas = alphas * weight[:, :, None]
+
     # outer left-multiply by alphas
     def lm_alphas(t):
         return torch.matmul(alphas[..., None], t).reshape(bs, n, 12)
@@ -82,9 +86,6 @@ def _build_M(y, alphas, weight):
         dim=-1,
     ).reshape(bs, -1, 12)
 
-    if weight is not None:
-        M = M * weight.repeat(1, 2)[:, :, None]
-
     return M
 
 

tests/test_perspective_n_points.py

@@ -24,6 +24,21 @@ class TestPerspectiveNPoints(TestCaseMixin, unittest.TestCase):
         super().setUp()
         torch.manual_seed(42)
 
+    @classmethod
+    def _generate_epnp_test_from_2d(cls, y):
+        """
+        Instantiate random x_world, x_cam, R, T given a set of input
+        2D projections y.
+        """
+        batch_size = y.shape[0]
+        x_cam = torch.cat((y, torch.rand_like(y[:, :, :1]) * 2.0 + 3.5), dim=2)
+        x_cam[:, :, :2] *= x_cam[:, :, 2:]  # unproject
+        R = rotation_conversions.random_rotations(batch_size).to(y)
+        T = torch.randn_like(R[:, :1, :])
+        T[:, :, 2] = (T[:, :, 2] + 3.0).clamp(2.0)
+        x_world = torch.matmul(x_cam - T, R.transpose(1, 2))
+        return x_cam, x_world, R, T
+
     def _run_and_print(self, x_world, y, R, T, print_stats, skip_q, check_output=False):
         sol = perspective_n_points.efficient_pnp(
             x_world, y.expand_as(x_world[:, :, :2]), skip_quadratic_eq=skip_q
@@ -45,16 +60,16 @@ class TestPerspectiveNPoints(TestCaseMixin, unittest.TestCase):
             )
 
             self.assertClose(err_2d, sol.err_2d, msg=assert_msg)
-            self.assertTrue((err_2d < 1e-4).all(), msg=assert_msg)
+            self.assertTrue((err_2d < 5e-4).all(), msg=assert_msg)
 
             def norm_fn(t):
                 return t.norm(dim=-1)
 
             self.assertNormsClose(
-                T, sol.T[:, None, :], rtol=3e-3, norm_fn=norm_fn, msg=assert_msg
+                T, sol.T[:, None, :], rtol=4e-3, norm_fn=norm_fn, msg=assert_msg
             )
             self.assertNormsClose(
-                R_quat, R_est_quat, rtol=3e-4, norm_fn=norm_fn, msg=assert_msg
+                R_quat, R_est_quat, rtol=3e-3, norm_fn=norm_fn, msg=assert_msg
             )
 
         if print_stats:
@@ -71,12 +86,9 @@ class TestPerspectiveNPoints(TestCaseMixin, unittest.TestCase):
                 print("T_hat | T_gt\n", T_gt)
 
     def _testcase_from_2d(self, y, print_stats, benchmark, skip_q=False):
-        x_cam = torch.cat((y, torch.rand_like(y[:, :1]) * 2.0 + 3.5), dim=1)
-        x_cam[:, :2] *= x_cam[:, 2:]  # unproject
-
-        R = rotation_conversions.random_rotations(16).to(y)
-        T = torch.randn_like(R[:, :1, :])
-        x_world = torch.matmul(x_cam - T, R.transpose(1, 2))
+        x_cam, x_world, R, T = TestPerspectiveNPoints._generate_epnp_test_from_2d(
+            y[None].repeat(16, 1, 1)
+        )
 
         if print_stats:
             print("Run without noise")
@@ -129,3 +141,45 @@ class TestPerspectiveNPoints(TestCaseMixin, unittest.TestCase):
                     benchmark=False,
                     skip_q=skip_q,
                 )
+
+    def test_weighted_perspective_n_points(self, batch_size=16, num_pts=200):
+        # instantiate random x_world and y
+        y = torch.randn((batch_size, num_pts, 2)).cuda() / 3.0
+        x_cam, x_world, R, T = TestPerspectiveNPoints._generate_epnp_test_from_2d(y)
+
+        # randomly drop 50% of the rows
+        weights = (torch.rand_like(x_world[:, :, 0]) > 0.5).float()
+
+        # make sure we retain at least 6 points for each case
+        weights[:, :6] = 1.0
+
+        # fill ignored y with trash to ensure that we get different
+        # solution in case the weighting is wrong
+        y = y + (1 - weights[:, :, None]) * 100.0
+
+        def norm_fn(t):
+            return t.norm(dim=-1)
+
+        for skip_quadratic_eq in (True, False):
+            # get the solution for the 0/1 weighted case
+            sol = perspective_n_points.efficient_pnp(
+                x_world, y, skip_quadratic_eq=skip_quadratic_eq, weights=weights
+            )
+            sol_R_quat = rotation_conversions.matrix_to_quaternion(sol.R)
+            sol_T = sol.T
+
+            # check that running only on points with non-zero weights ends in the
+            # same place as running the 0/1 weighted version
+            for i in range(batch_size):
+                ok = weights[i] > 0
+                x_world_ok = x_world[i, ok][None]
+                y_ok = y[i, ok][None]
+                sol_ok = perspective_n_points.efficient_pnp(
+                    x_world_ok, y_ok, skip_quadratic_eq=False
+                )
+                R_est_quat_ok = rotation_conversions.matrix_to_quaternion(sol_ok.R)
+
+                self.assertNormsClose(sol_T[i], sol_ok.T[0], rtol=3e-3, norm_fn=norm_fn)
+                self.assertNormsClose(
+                    sol_R_quat[i], R_est_quat_ok[0], rtol=3e-4, norm_fn=norm_fn
+                )