Flip height and width in the detector (#357)

* Flip height and width; fix reorientations

* Fix orientations

* Rename f to sdd

* Handle errors in volume masking

* Fix perspective projection

* Fix x/y conventions

* Bump version

Author: eigenvivek

diffdrr/__init__.py

@@ -1 +1 @@
-__version__ = "0.4.5"
+__version__ = "0.4.6"

diffdrr/data.py

@@ -85,35 +85,37 @@ def read(
     # Frame-of-reference change
     if orientation == "AP":
         # Rotates the C-arm about the x-axis by 90 degrees
-        # Rotates the C-arm about the z-axis by -90 degrees
         reorient = torch.tensor(
             [
-                [0.0, 1.0, 0.0, 0.0],
-                [0.0, 0.0, -1.0, 0.0],
-                [-1.0, 0.0, 0.0, 0.0],
-                [0.0, 0.0, 0.0, 1.0],
-            ]
+                [1, 0, 0, 0],
+                [0, 0, -1, 0],
+                [0, 1, 0, 0],
+                [0, 0, 0, 1],
+            ],
+            dtype=torch.float32,
         )
     elif orientation == "PA":
         # Rotates the C-arm about the x-axis by 90 degrees
-        # Rotates the C-arm about the z-axis by 90 degrees
+        # Reverses the direction of the y-axis
         reorient = torch.tensor(
             [
-                [0.0, 1.0, 0.0, 0.0],
-                [0.0, 0.0, 1.0, 0.0],
-                [-1.0, 0.0, 0.0, 0.0],
-                [0.0, 0.0, 0.0, 1.0],
-            ]
+                [1, 0, 0, 0],
+                [0, 0, 1, 0],
+                [0, 1, 0, 0],
+                [0, 0, 0, 1],
+            ],
+            dtype=torch.float32,
         )
     elif orientation is None:
         # Identity transform
         reorient = torch.tensor(
             [
-                [1.0, 0.0, 0.0, 0.0],
-                [0.0, 1.0, 0.0, 0.0],
-                [0.0, 0.0, 1.0, 0.0],
-                [0.0, 0.0, 0.0, 1.0],
-            ]
+                [1, 0, 0, 0],
+                [0, 1, 0, 0],
+                [0, 0, 1, 0],
+                [0, 0, 0, 1],
+            ],
+            dtype=torch.float32,
         )
     else:
         raise ValueError(f"Unrecognized orientation {orientation}")
@@ -122,6 +124,7 @@ def read(
     subject = Subject(
         volume=volume,
         mask=mask,
+        orientation=orientation,
         reorient=reorient,
         density=density,
         fiducials=fiducials,
@@ -161,9 +164,13 @@ def read(
                 dim=0,
             )
 
-        subject.volume.data = subject.volume.data * mask
-        subject.mask.data = subject.mask.data * mask
-        subject.density.data = subject.density.data * mask
+        # Mask all volumes, unless error, then just mask the density
+        try:
+            subject.volume.data = subject.volume.data * mask
+            subject.mask.data = subject.mask.data * mask
+            subject.density.data = subject.density.data * mask
+        except:
+            subject.density.data = subject.density.data * mask
 
     return subject
 

diffdrr/detector.py

@@ -51,8 +51,8 @@ def __init__(
             "_calibration",
             torch.tensor(
                 [
-                    [dely, 0, 0, -y0],
-                    [0, delx, 0, -x0],
+                    [delx, 0, 0, x0],
+                    [0, dely, 0, y0],
                     [0, 0, sdd, 0],
                     [0, 0, 0, 1],
                 ]
@@ -65,19 +65,19 @@ def sdd(self):
 
     @property
     def delx(self):
-        return self._calibration[1, 1].item()
+        return self._calibration[0, 0].item()
 
     @property
     def dely(self):
-        return self._calibration[0, 0].item()
+        return self._calibration[1, 1].item()
 
     @property
     def x0(self):
-        return -self._calibration[1, -1].item()
+        return -self._calibration[0, -1].item()
 
     @property
     def y0(self):
-        return -self._calibration[0, -1].item()
+        return -self._calibration[1, -1].item()
 
     @property
     def reorient(self):
@@ -107,7 +107,7 @@ def _initialize_carm(self: Detector):
     center = torch.tensor([[0.0, 0.0, 1.0]], device=device)
 
     # Use the standard basis for the detector plane
-    basis = torch.tensor([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], device=device)
+    basis = torch.tensor([[0.0, 1.0, 0.0], [1.0, 0.0, 0.0]], device=device)
 
     # Construct the detector plane with different offsets for even or odd heights
     # These ensure that the detector plane is centered around (0, 0, 1)
@@ -117,8 +117,12 @@ def _initialize_carm(self: Detector):
     # Construct equally spaced points along the basis vectors
     t = torch.arange(-self.height // 2, self.height // 2, device=device) + h_off
     s = torch.arange(-self.width // 2, self.width // 2, device=device) + w_off
-    if self.reverse_x_axis:
+
+    t = -t
+    s = -s
+    if not self.reverse_x_axis:
         s = -s
+
     coefs = torch.cartesian_prod(t, s).reshape(-1, 2)
     target = torch.einsum("cd,nc->nd", basis, coefs)
     target += center

diffdrr/drr.py

@@ -230,8 +230,8 @@ def set_intrinsics_(
         width if width is not None else self.detector.width,
         delx if delx is not None else self.detector.delx,
         dely if dely is not None else self.detector.dely,
-        x0 if x0 is not None else self.detector.x0,
-        y0 if y0 is not None else self.detector.y0,
+        x0 if x0 is not None else -self.detector.x0,
+        y0 if y0 is not None else -self.detector.y0,
         self.subject.reorient,
         n_subsample if n_subsample is not None else self.detector.n_subsample,
         reverse_x_axis if reverse_x_axis is not None else self.detector.reverse_x_axis,
@@ -256,14 +256,21 @@ def perspective_projection(
     pts: torch.Tensor,
 ):
     """Project points in world coordinates (3D) onto the pixel plane (2D)."""
+    # Poses in DiffDRR are world2camera, but perspective transforms use camera2world, so invert
     extrinsic = (self.detector.reorient.compose(pose)).inverse()
     x = extrinsic(pts)
+
+    # Project onto the detector plane
     x = torch.einsum("ij, bnj -> bni", self.detector.intrinsic, x)
     z = x[..., -1].unsqueeze(-1).clone()
     x = x / z
+
+    # Move origin to upper-left corner
+    x[..., 1] = self.detector.height - x[..., 1]
     if self.detector.reverse_x_axis:
-        x[..., 1] = self.detector.width - x[..., 1]
-    return x[..., :2].flip(-1)
+        x[..., 0] = self.detector.width - x[..., 0]
+
+    return x[..., :2]
 
 # %% ../notebooks/api/00_drr.ipynb 14
 from torch.nn.functional import pad
@@ -276,7 +283,7 @@ def inverse_projection(
     pts: torch.Tensor,
 ):
     """Backproject points in pixel plane (2D) onto the image plane in world coordinates (3D)."""
-    pts = pts.flip(-1)
+    # pts = pts.flip(-1)
     if self.detector.reverse_x_axis:
         pts[..., 1] = self.detector.width - pts[..., 1]
     x = self.detector.sdd * torch.einsum(

diffdrr/utils.py

@@ -54,6 +54,7 @@ def resample(
 
 # %% ../notebooks/api/07_utils.ipynb 6
 from kornia.geometry.camera.pinhole import PinholeCamera as KorniaPinholeCamera
+from torchio import Subject
 
 from diffdrr.detector import Detector
 
@@ -66,9 +67,11 @@ def __init__(
         height: torch.Tensor,
         width: torch.Tensor,
         detector: Detector,
+        subject: Subject,
     ):
         super().__init__(intrinsics, extrinsics, height, width)
-        self.f = detector.sdd
+        multiplier = -1 if subject.orientation == "PA" else 1
+        self.sdd = multiplier * detector.sdd
         self.delx = detector.delx
         self.dely = detector.dely
         self.x0 = detector.x0
@@ -94,9 +97,9 @@ def pose(self):
 
 from kornia.geometry.calibration import solve_pnp_dlt
 
+from .detector import make_intrinsic_matrix
 from .drr import DRR
 from .pose import RigidTransform
-from .detector import make_intrinsic_matrix
 
 
 def get_pinhole_camera(
@@ -107,14 +110,12 @@ def get_pinhole_camera(
     pose = deepcopy(pose).to(device="cpu", dtype=dtype)
 
     # Make the intrinsic matrix (in pixels)
-    fx = drr.detector.sdd / drr.detector.delx
-    fy = drr.detector.sdd / drr.detector.dely
+    multiplier = -1 if drr.subject.orientation == "PA" else 1
+    fx = multiplier * drr.detector.sdd / drr.detector.delx
+    fy = multiplier * drr.detector.sdd / drr.detector.dely
     u0 = drr.detector.x0 / drr.detector.delx + drr.detector.width / 2
     v0 = drr.detector.y0 / drr.detector.dely + drr.detector.height / 2
-    intrinsics = torch.eye(4)[None]
-    intrinsics[0, :3, :3] = make_intrinsic_matrix(drr.detector)
-
-    torch.tensor(
+    intrinsics = torch.tensor(
         [
             [
                 [fx, 0.0, u0, 0.0],
@@ -156,6 +157,7 @@ def get_pinhole_camera(
         torch.tensor([drr.detector.height]),
         torch.tensor([drr.detector.width]),
         drr.detector,
+        drr.subject,
     )
 
     return camera

notebooks/api/00_drr.ipynb

@@ -357,8 +357,8 @@
     "        width if width is not None else self.detector.width,\n",
     "        delx if delx is not None else self.detector.delx,\n",
     "        dely if dely is not None else self.detector.dely,\n",
-    "        x0 if x0 is not None else self.detector.x0,\n",
-    "        y0 if y0 is not None else self.detector.y0,\n",
+    "        x0 if x0 is not None else -self.detector.x0,\n",
+    "        y0 if y0 is not None else -self.detector.y0,\n",
     "        self.subject.reorient,\n",
     "        n_subsample if n_subsample is not None else self.detector.n_subsample,\n",
     "        reverse_x_axis if reverse_x_axis is not None else self.detector.reverse_x_axis,\n",
@@ -399,14 +399,21 @@
     "    pts: torch.Tensor,\n",
     "):\n",
     "    \"\"\"Project points in world coordinates (3D) onto the pixel plane (2D).\"\"\"\n",
+    "    # Poses in DiffDRR are world2camera, but perspective transforms use camera2world, so invert\n",
     "    extrinsic = (self.detector.reorient.compose(pose)).inverse()\n",
     "    x = extrinsic(pts)\n",
+    "\n",
+    "    # Project onto the detector plane\n",
     "    x = torch.einsum(\"ij, bnj -> bni\", self.detector.intrinsic, x)\n",
     "    z = x[..., -1].unsqueeze(-1).clone()\n",
     "    x = x / z\n",
+    "\n",
+    "    # Move origin to upper-left corner\n",
+    "    x[..., 1] = self.detector.height - x[..., 1]\n",
     "    if self.detector.reverse_x_axis:\n",
-    "        x[..., 1] = self.detector.width - x[..., 1]\n",
-    "    return x[..., :2].flip(-1)"
+    "        x[..., 0] = self.detector.width - x[..., 0]\n",
+    "    \n",
+    "    return x[..., :2]"
    ]
   },
   {
@@ -427,7 +434,7 @@
     "    pts: torch.Tensor,\n",
     "):\n",
     "    \"\"\"Backproject points in pixel plane (2D) onto the image plane in world coordinates (3D).\"\"\"\n",
-    "    pts = pts.flip(-1)\n",
+    "    # pts = pts.flip(-1)\n",
     "    if self.detector.reverse_x_axis:\n",
     "        pts[..., 1] = self.detector.width - pts[..., 1]\n",
     "    x = self.detector.sdd * torch.einsum(\n",

notebooks/api/02_detector.ipynb

@@ -106,8 +106,8 @@
     "            \"_calibration\",\n",
     "            torch.tensor(\n",
     "                [\n",
-    "                    [dely, 0, 0, -y0],\n",
-    "                    [0, delx, 0, -x0],\n",
+    "                    [delx, 0, 0, x0],\n",
+    "                    [0, dely, 0, y0],\n",
     "                    [0, 0, sdd, 0],\n",
     "                    [0, 0, 0, 1],\n",
     "                ]\n",
@@ -120,19 +120,19 @@
     "\n",
     "    @property\n",
     "    def delx(self):\n",
-    "        return self._calibration[1, 1].item()\n",
+    "        return self._calibration[0, 0].item()\n",
     "\n",
     "    @property\n",
     "    def dely(self):\n",
-    "        return self._calibration[0, 0].item()\n",
+    "        return self._calibration[1, 1].item()\n",
     "\n",
     "    @property\n",
     "    def x0(self):\n",
-    "        return -self._calibration[1, -1].item()\n",
+    "        return -self._calibration[0, -1].item()\n",
     "\n",
     "    @property\n",
     "    def y0(self):\n",
-    "        return -self._calibration[0, -1].item()\n",
+    "        return -self._calibration[1, -1].item()\n",
     "\n",
     "    @property\n",
     "    def reorient(self):\n",
@@ -170,7 +170,7 @@
     "    center = torch.tensor([[0.0, 0.0, 1.0]], device=device)\n",
     "\n",
     "    # Use the standard basis for the detector plane\n",
-    "    basis = torch.tensor([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], device=device)\n",
+    "    basis = torch.tensor([[0.0, 1.0, 0.0], [1.0, 0.0, 0.0]], device=device)\n",
     "\n",
     "    # Construct the detector plane with different offsets for even or odd heights\n",
     "    # These ensure that the detector plane is centered around (0, 0, 1)\n",
@@ -180,8 +180,12 @@
     "    # Construct equally spaced points along the basis vectors\n",
     "    t = torch.arange(-self.height // 2, self.height // 2, device=device) + h_off\n",
     "    s = torch.arange(-self.width // 2, self.width // 2, device=device) + w_off\n",
-    "    if self.reverse_x_axis:\n",
+    "\n",
+    "    t = -t\n",
+    "    s = -s\n",
+    "    if not self.reverse_x_axis:\n",
     "        s = -s\n",
+    "\n",
     "    coefs = torch.cartesian_prod(t, s).reshape(-1, 2)\n",
     "    target = torch.einsum(\"cd,nc->nd\", basis, coefs)\n",
     "    target += center\n",