fixed recovery of metric depth for orthographic cameras

pyrender/camera.py

@@ -91,6 +91,20 @@ def get_projection_matrix(self, width=None, height=None):
         """
         pass
 
+    @abc.abstractmethod
+    def inverse_transform_depth_ndc(self, depth_ndc):
+        """
+        Unproject the depth in Normalized Device Coordinates
+        to metric values in eye (camera) reference frame.
+        :param
+            depth_ndc: np.array np.float32
+                depth in [-1, 1] in normalized device coordinates
+        :return:
+            depth: np.array np.float32
+                depth in metric units in eye (camera) reference frame.
+        """
+        pass
+
 
 class PerspectiveCamera(Camera):
 
@@ -205,6 +219,19 @@ def get_projection_matrix(self, width=None, height=None):
 
         return P
 
+    def inverse_transform_depth_ndc(self, depth_ndc):
+        inf_inds = (depth_ndc == 1)
+        noninf = np.logical_not(inf_inds)
+        depth_img = depth_ndc
+        if self.zfar is None:
+            depth_img[noninf] = 2 * self.znear / (1.0 - depth_img[noninf])
+        else:
+            depth_img[noninf] = ((2.0 * self.znear * self.zfar) /
+                                 (self.zfar + self.znear - depth_img[noninf] *
+                                  (self.zfar - self.znear)))
+        depth_img[inf_inds] = 0.0
+        return depth_img
+
 
 class OrthographicCamera(Camera):
     """A perspective camera for perspective projection.
@@ -309,6 +336,17 @@ def get_projection_matrix(self, width=None, height=None):
         P[3][3] = 1.0
         return P
 
+    def inverse_transform_depth_ndc(self, depth_ndc):
+        inf_inds = (depth_ndc == 1)
+        noninf = np.logical_not(inf_inds)
+        depth_img = depth_ndc
+        if self.zfar is None:
+            depth_img[noninf] = 2 * self.znear / (1.0 - depth_img[noninf])
+        else:
+            depth_img[noninf] = (depth_img[noninf] * (self.zfar - self.znear) + self.zfar + self.znear) / 2.0
+        depth_img[inf_inds] = 0.0
+        return depth_img
+
 
 class IntrinsicsCamera(Camera):
     """A perspective camera with custom intrinsics.
@@ -432,6 +470,19 @@ def get_projection_matrix(self, width, height):
 
         return P
 
+    def inverse_transform_depth_ndc(self, depth_ndc):
+        inf_inds = (depth_ndc == 1)
+        noninf = np.logical_not(inf_inds)
+        depth_img = depth_ndc
+        if self.zfar is None:
+            depth_img[noninf] = 2 * self.znear / (1.0 - depth_img[noninf])
+        else:
+            depth_img[noninf] = ((2.0 * self.znear * self.zfar) /
+                                 (self.zfar + self.znear - depth_img[noninf] *
+                                  (self.zfar - self.znear)))
+        depth_img[inf_inds] = 0.0
+        return depth_img
+
 
 __all__ = ['Camera', 'PerspectiveCamera', 'OrthographicCamera',
            'IntrinsicsCamera']

pyrender/renderer.py

@@ -1117,18 +1117,8 @@ def _read_main_framebuffer(self, scene, flags):
         depth_im = np.frombuffer(depth_buf, dtype=np.float32)
         depth_im = depth_im.reshape((height, width))
         depth_im = np.flip(depth_im, axis=0)
-        inf_inds = (depth_im == 1.0)
         depth_im = 2.0 * depth_im - 1.0
-        z_near = scene.main_camera_node.camera.znear
-        z_far = scene.main_camera_node.camera.zfar
-        noninf = np.logical_not(inf_inds)
-        if z_far is None:
-            depth_im[noninf] = 2 * z_near / (1.0 - depth_im[noninf])
-        else:
-            depth_im[noninf] = ((2.0 * z_near * z_far) /
-                                (z_far + z_near - depth_im[noninf] *
-                                (z_far - z_near)))
-        depth_im[inf_inds] = 0.0
+        depth_im = scene.main_camera_node.camera.inverse_transform_depth_ndc(depth_im)
 
         # Resize for macos if needed
         if sys.platform == 'darwin':