Fixes for RayBundle plotting

Summary:
Fixes some issues with RayBundle plotting:
- allows plotting raybundles on gpu
- view -> reshape since we do not require contiguous raybundle tensors as input

Reviewed By: bottler, shapovalov

Differential Revision: D42665923

fbshipit-source-id: e9c6c7810428365dca4cb5ec80ef15ff28644163

Author: davnov134

pytorch3d/vis/__init__.py

@@ -3,3 +3,19 @@
 #
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
+
+import warnings
+
+
+try:
+    from .plotly_vis import get_camera_wireframe, plot_batch_individually, plot_scene
+except ModuleNotFoundError as err:
+    if "plotly" in str(err):
+        warnings.warn(
+            "Cannot import plotly-based visualization code."
+            " Please install plotly to enable (pip install plotly)."
+        )
+    else:
+        raise
+
+from .texture_vis import texturesuv_image_matplotlib, texturesuv_image_PIL

pytorch3d/vis/plotly_vis.py

@@ -100,6 +100,7 @@ class Lighting(NamedTuple):  # pragma: no cover
     vertexnormalsepsilon: float = 1e-12
 
 
+@torch.no_grad()
 def plot_scene(
     plots: Dict[str, Dict[str, Struct]],
     *,
@@ -407,6 +408,7 @@ def plot_scene(
     return fig
 
 
+@torch.no_grad()
 def plot_batch_individually(
     batched_structs: Union[
         List[Struct],
@@ -888,8 +890,12 @@ def _add_ray_bundle_trace(
     )
 
     # make the ray lines for plotly plotting
-    nan_tensor = torch.Tensor([[float("NaN")] * 3])
-    ray_lines = torch.empty(size=(1, 3))
+    nan_tensor = torch.tensor(
+        [[float("NaN")] * 3],
+        device=ray_lines_endpoints.device,
+        dtype=ray_lines_endpoints.dtype,
+    )
+    ray_lines = torch.empty(size=(1, 3), device=ray_lines_endpoints.device)
     for ray_line in ray_lines_endpoints:
         # We combine the ray lines into a single tensor to plot them in a
         # single trace. The NaNs are inserted between sets of ray lines
@@ -952,7 +958,7 @@ def _add_ray_bundle_trace(
     current_layout = fig["layout"][plot_scene]
 
     # update the bounds of the axes for the current trace
-    all_ray_points = ray_bundle_to_ray_points(ray_bundle).view(-1, 3)
+    all_ray_points = ray_bundle_to_ray_points(ray_bundle).reshape(-1, 3)
     ray_points_center = all_ray_points.mean(dim=0)
     max_expand = (all_ray_points.max(0)[0] - all_ray_points.min(0)[0]).max().item()
     _update_axes_bounds(ray_points_center, float(max_expand), current_layout)
@@ -1002,6 +1008,7 @@ def _update_axes_bounds(
         max_expand: the maximum spread in any dimension of the trace's vertices.
         current_layout: the plotly figure layout scene corresponding to the referenced trace.
     """
+    verts_center = verts_center.detach().cpu()
     verts_min = verts_center - max_expand
     verts_max = verts_center + max_expand
     bounds = torch.t(torch.stack((verts_min, verts_max)))

tests/test_vis.py

@@ -0,0 +1,74 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import unittest
+
+import torch
+from pytorch3d.renderer import HeterogeneousRayBundle, PerspectiveCameras, RayBundle
+from pytorch3d.structures import Meshes, Pointclouds
+from pytorch3d.transforms import random_rotations
+
+# Some of these imports are only needed for testing code coverage
+from pytorch3d.vis import (  # noqa: F401
+    get_camera_wireframe,  # noqa: F401
+    plot_batch_individually,  # noqa: F401
+    plot_scene,
+    texturesuv_image_PIL,  # noqa: F401
+)
+
+
+class TestPlotlyVis(unittest.TestCase):
+    def test_plot_scene(
+        self,
+        B: int = 3,
+        n_rays: int = 128,
+        n_pts_per_ray: int = 32,
+        n_verts: int = 32,
+        n_edges: int = 64,
+        n_pts: int = 256,
+    ):
+        """
+        Tests plotting of all supported structures using plot_scene.
+        """
+        for device in ["cpu", "cuda:0"]:
+            plot_scene(
+                {
+                    "scene": {
+                        "ray_bundle": RayBundle(
+                            origins=torch.randn(B, n_rays, 3, device=device),
+                            xys=torch.randn(B, n_rays, 2, device=device),
+                            directions=torch.randn(B, n_rays, 3, device=device),
+                            lengths=torch.randn(
+                                B, n_rays, n_pts_per_ray, device=device
+                            ),
+                        ),
+                        "heterogeneous_ray_bundle": HeterogeneousRayBundle(
+                            origins=torch.randn(B * n_rays, 3, device=device),
+                            xys=torch.randn(B * n_rays, 2, device=device),
+                            directions=torch.randn(B * n_rays, 3, device=device),
+                            lengths=torch.randn(
+                                B * n_rays, n_pts_per_ray, device=device
+                            ),
+                            camera_ids=torch.randint(
+                                low=0, high=B, size=(B * n_rays,), device=device
+                            ),
+                        ),
+                        "camera": PerspectiveCameras(
+                            R=random_rotations(B, device=device),
+                            T=torch.randn(B, 3, device=device),
+                        ),
+                        "mesh": Meshes(
+                            verts=torch.randn(B, n_verts, 3, device=device),
+                            faces=torch.randint(
+                                low=0, high=n_verts, size=(B, n_edges, 3), device=device
+                            ),
+                        ),
+                        "point_clouds": Pointclouds(
+                            points=torch.randn(B, n_pts, 3, device=device),
+                        ),
+                    }
+                }
+            )