Improve Documentation - Fix code links and improve tutorial layout (#…

…6321) * Correct code tagging in interactive_visualization tutorial. * Highlight the methods that are being introduced at the start. * Non-Blocking VIZ - Add add_geometry step which is not optional. * Non-Blocking VIZ - Split main code to simpler function and fix code blocks in docs. * Fix Typos * Update docs with new argument values. * Fix grammar and dedent for overly indented code * Integrate_Custom - Optimize imports, add early return, use f-strings and remove unnecessary parentheses. * Fix default value and remove BooleanOptionalAction, its only available >= python 3.9 * Disable use of cuda for tensor based reconstruction * Customized Integration - Fix code lines * Ray Casting - Fix code lines * Dense Slam - Fix code lines and slight grammar. * Fix some f-strings and extra parantheses * Make Fragments - Fix code lines * Code for batch processing was removed long back in #521 in 2018 * Register Fragments - Fix code docs * Register fragments - Fix code lines * Integrate Scene - Fix code lines * Change layout of tutorials - only done at top level. * Add new lines before bullet/numbered list, else list rendering fails. * Bugfix - Fix code lines for Integration and add filename as done in v0.14.0 * Remove python version 3.6 as supported versions by open3d. * Fix line number using lineno-match directive. * Fix some typos detected automatically. * Replace prepend with lineno-match * Correct filename documented in the files * Fix lineno-start. * Fix PR comments * Revert "Code for batch processing was removed long back in #521 in 2018" This reverts commit d65fb67. * Correct batch processing section documentation and native multiprocessing. * Use native multiprocessing for register fragments and refinement of registration --------- Co-authored-by: Sameer Sheorey <41028320+ssheorey@users.noreply.github.com>
isl-org · Sep 26, 2023 · f1a0f3e · f1a0f3e
1 parent b2f5cef
commit f1a0f3e
Show file tree

Hide file tree

Showing 40 changed files with 359 additions and 310 deletions.
diff --git a/README.md b/README.md
@@ -45,7 +45,7 @@ For more, please visit the [Open3D documentation](http://www.open3d.org/docs).
 ## Python quick start
 
 Pre-built pip packages support Ubuntu 18.04+, macOS 10.15+ and Windows 10+
-(64-bit) with Python 3.6-3.10.
+(64-bit) with Python 3.7-3.10.
 
 ```bash
 # Install

diff --git a/cpp/open3d/core/AdvancedIndexing.cpp b/cpp/open3d/core/AdvancedIndexing.cpp
@@ -206,7 +206,7 @@ void AdvancedIndexPreprocessor::RunPreprocess() {
     // If the indexed_shape_ contains a dimension of size 0 but the
     // replacement shape does not, the index is out of bounds. This is because
     // there is no valid number to index an empty tensor.
-    // Normally, out of bounds is detected in the advanded indexing kernel. We
+    // Normally, out of bounds is detected in the advanced indexing kernel. We
     // detected here for more helpful error message.
     auto contains_zero = [](const SizeVector& vals) -> bool {
         return std::any_of(vals.begin(), vals.end(),

diff --git a/cpp/open3d/core/AdvancedIndexing.h b/cpp/open3d/core/AdvancedIndexing.h
@@ -49,7 +49,7 @@ class AdvancedIndexPreprocessor {
             const Tensor& tensor, const std::vector<Tensor>& index_tensors);
 
     /// Expand all tensors to the broadcasted shape, 0-dim tensors are ignored.
-    /// Thorws exception if the common broadcasted shape does not exist.
+    /// Throws exception if the common broadcasted shape does not exist.
     static std::pair<std::vector<Tensor>, SizeVector>
     ExpandToCommonShapeExceptZeroDim(const std::vector<Tensor>& index_tensors);
 
@@ -127,7 +127,7 @@ class AdvancedIndexer {
         if (indexed_shape.size() != indexed_strides.size()) {
             utility::LogError(
                     "Internal error: indexed_shape's ndim {} does not equal to "
-                    "indexd_strides' ndim {}",
+                    "indexed_strides' ndim {}",
                     indexed_shape.size(), indexed_strides.size());
         }
         num_indices_ = indexed_shape.size();

diff --git a/cpp/open3d/geometry/BoundingVolume.h b/cpp/open3d/geometry/BoundingVolume.h
@@ -21,7 +21,7 @@ class AxisAlignedBoundingBox;
 /// \brief A bounding box oriented along an arbitrary frame of reference.
 ///
 /// The oriented bounding box is defined by its center position, rotation
-/// maxtrix and extent.
+/// matrix and extent.
 class OrientedBoundingBox : public Geometry3D {
 public:
     /// \brief Default constructor.

diff --git a/cpp/open3d/geometry/VoxelGrid.h b/cpp/open3d/geometry/VoxelGrid.h
@@ -245,7 +245,7 @@ class VoxelGrid : public Geometry3D {
 public:
     /// Size of the voxel.
     double voxel_size_ = 0.0;
-    /// Coorindate of the origin point.
+    /// Coordinate of the origin point.
     Eigen::Vector3d origin_ = Eigen::Vector3d::Zero();
     /// Voxels contained in voxel grid
     std::unordered_map<Eigen::Vector3i,

diff --git a/cpp/open3d/ml/impl/continuous_conv/ContinuousConvBackpropFilter.cuh b/cpp/open3d/ml/impl/continuous_conv/ContinuousConvBackpropFilter.cuh
@@ -45,7 +45,7 @@ namespace impl {
 /// \param texture_alignment    The texture alignment in bytes. This is used
 ///        for allocating segments within the temporary memory.
 ///
-/// \param filter_backrop    Output array for the computed filter gradient
+/// \param filter_backprop    Output array for the computed filter gradient
 ///        with shape [depth,height,width, inp channels, out channels]
 ///
 /// \param filter_dims    The sizes of the filter dimensions. The size of

diff --git a/cpp/open3d/ml/tensorflow/misc/BuildSpatialHashTableOps.cpp b/cpp/open3d/ml/tensorflow/misc/BuildSpatialHashTableOps.cpp
@@ -78,7 +78,7 @@ The following example shows how **build_spatial_hash_table** and
 
   radius = 1.0
 
-  # build the spatial hash table for fixex_radius_search
+  # build the spatial hash table for fixed_radius_search
   table = ml3d.ops.build_spatial_hash_table(points,
                                             radius,
                                             points_row_splits=torch.LongTensor([0,5]),

diff --git a/cpp/open3d/t/geometry/BoundingVolume.cpp b/cpp/open3d/t/geometry/BoundingVolume.cpp
@@ -318,7 +318,7 @@ OrientedBoundingBox::OrientedBoundingBox(const core::Tensor &center,
         utility::LogError(
                 "Invalid oriented bounding box. Please make sure the values of "
                 "extent are all positive and the rotation matrix is "
-                "othogonal.");
+                "orthogonal.");
     }
 }
 

diff --git a/cpp/open3d/t/geometry/BoundingVolume.h b/cpp/open3d/t/geometry/BoundingVolume.h
@@ -146,7 +146,7 @@ class AxisAlignedBoundingBox : public Geometry, public DrawableGeometry {
             const utility::optional<core::Tensor> &center = utility::nullopt);
 
     /// \brief Add operation for axis-aligned bounding box.
-    /// The device of ohter box must be the same as the device of the current
+    /// The device of other box must be the same as the device of the current
     /// box.
     AxisAlignedBoundingBox &operator+=(const AxisAlignedBoundingBox &other);
 
@@ -223,7 +223,7 @@ class AxisAlignedBoundingBox : public Geometry, public DrawableGeometry {
 /// \brief A bounding box oriented along an arbitrary frame of reference.
 ///
 /// - (center, rotation, extent): The oriented bounding box is defined by its
-/// center position, rotation maxtrix and extent.
+/// center position, rotation matrix and extent.
 ///     - Usage
 ///         - OrientedBoundingBox::GetCenter()
 ///         - OrientedBoundingBox::SetCenter(const core::Tensor &center)

diff --git a/cpp/open3d/visualization/gui/Window.cpp b/cpp/open3d/visualization/gui/Window.cpp
@@ -707,7 +707,7 @@ Widget::DrawResult DrawChild(DrawContext& dc,
                              const char* name,
                              std::shared_ptr<Widget> child,
                              Mode mode) {
-    // Note: ImGUI's concept of a "window" is really a moveable child of the
+    // Note: ImGUI's concept of a "window" is really a movable child of the
     //       OS window. We want a child to act like a child of the OS window,
     //       like native UI toolkits, Qt, etc. So the top-level widgets of
     //       a window are drawn using ImGui windows whose frame is specified

diff --git a/docs/index.rst b/docs/index.rst
@@ -34,13 +34,13 @@ Open3D: A Modern Library for 3D Data Processing
     :maxdepth: 2
     :caption: Tutorial
 
+    tutorial/core/index
     tutorial/geometry/index
     tutorial/t_geometry/index
+    tutorial/data/index
+    tutorial/visualization/index
     tutorial/pipelines/index
     tutorial/t_pipelines/index
-    tutorial/visualization/index
-    tutorial/core/index
-    tutorial/data/index
     tutorial/reconstruction_system/index
     tutorial/t_reconstruction_system/index
     tutorial/sensor/index

diff --git a/docs/jupyter/core/hashmap.ipynb b/docs/jupyter/core/hashmap.ipynb
@@ -88,19 +88,23 @@
     "Next we show how to insert a batch of (key, value) pairs. You'll need to prepare two tensors:\n",
     "\n",
     "The `keys` tensor contains all keys. \n",
+    "\n",
     "- The `keys` tensor must be on the same device as the hash map. \n",
     "- The shape of the `keys` tensor is `key_elment_shape` with `N` prefixed to the front. \n",
     "\n",
     "For example \n",
+    " \n",
     "1. if `key_element_shape == ()`, `keys.shape == (N,)`; \n",
     "2. if `key_element_shape == (3,)`, `keys.shape == (N, 3).`; \n",
     "3. if `key_element_shape == (8, 8, 8)`, `keys.shape == (N, 8, 8, 8).`\n",
     "    \n",
     "The `vals` tensor contains all values. \n",
+    " \n",
     "- The `vals` tensor must be on the same device as the hash map. \n",
     "- The shape of the `vals` tensor is `val_elment_shape` with `N` prefixed to the front. \n",
     "\n",
     "For example \n",
+    "\n",
     "1. if `val_elment_shape == ()`, `vals.shape == (N,)`; \n",
     "2. if `val_elment_shape == (3,)`, `vals.shape == (N, 3).`;\n",
     "3. if `val_elment_shape == (8, 8, 8)`, `vals.shape == (N, 8, 8, 8).`"

diff --git a/docs/tutorial/reconstruction_system/integrate_scene.rst b/docs/tutorial/reconstruction_system/integrate_scene.rst
@@ -21,9 +21,10 @@ Integrate RGBD frames
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/integrate_scene.py
    :language: python
-   :lineno-start: 38
-   :lines: 27,40-72
+   :pyobject: scalable_integrate_rgb_frames
+   :end-at:        o3d.visualization.draw_geometries([mesh])
    :linenos:
+   :lineno-match:
 
 This function first reads the alignment results from both
 :ref:`reconstruction_system_make_fragments` and

diff --git a/docs/tutorial/reconstruction_system/make_fragments.rst b/docs/tutorial/reconstruction_system/make_fragments.rst
@@ -25,9 +25,9 @@ Register RGBD image pairs
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/make_fragments.py
    :language: python
-   :lineno-start: 46
-   :lines: 27,47-76
+   :pyobject: register_one_rgbd_pair
    :linenos:
+   :lineno-match:
 
 The function reads a pair of RGBD images and registers the ``source_rgbd_image``
 to the ``target_rgbd_image``. The Open3D function ``compute_rgbd_odometry`` is
@@ -45,9 +45,9 @@ Multiway registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/make_fragments.py
    :language: python
-   :lineno-start: 76
-   :lines: 27,77-123
+   :pyobject: make_posegraph_for_fragment
    :linenos:
+   :lineno-match:
 
 This script uses the technique demonstrated in
 :ref:`/tutorial/pipelines/multiway_registration.ipynb`. The function
@@ -61,9 +61,9 @@ function ``optimize_posegraph_for_fragment``.
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/optimize_posegraph.py
    :language: python
-   :lineno-start: 51
-   :lines: 27,52-63
+   :pyobject: optimize_posegraph_for_fragment
    :linenos:
+   :lineno-match:
 
 This function calls ``global_optimization`` to estimate poses of the RGBD images.
 
@@ -74,23 +74,28 @@ Make a fragment
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/make_fragments.py
    :language: python
-   :lineno-start: 124
-   :lines: 27,125-146
+   :pyobject: integrate_rgb_frames_for_fragment
    :linenos:
+   :lineno-match:
 
-Once the poses are estimates, :ref:`/tutorial/pipelines/rgbd_integration.ipynb`
+Once the poses are estimated, :ref:`/tutorial/pipelines/rgbd_integration.ipynb`
 is used to reconstruct a colored fragment from each RGBD sequence.
 
 Batch processing
 ``````````````````````````````````````
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/make_fragments.py
    :language: python
-   :lineno-start: 181
-   :lines: 27,182-205
+   :start-at: def process_single_fragment(fragment_id, color_files, depth_files, n_files,
    :linenos:
+   :lineno-match:
+
+The ``process_single_fragment`` function calls each individual function explained above.
+The ``run`` function determines the number of fragments to generate based on the number
+of images in the dataset and the configuration value ``n_frames_per_fragment``.
+Subsequently, it invokes ``process_single_fragment`` for each of these fragments.
+Furthermore, it leverages multiprocessing to speed up computation of all fragments.
 
-The main function calls each individual function explained above.
 
 .. _reconstruction_system_make_fragments_results:
 

diff --git a/docs/tutorial/reconstruction_system/refine_registration.rst b/docs/tutorial/reconstruction_system/refine_registration.rst
@@ -20,9 +20,9 @@ Fine-grained registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/refine_registration.py
    :language: python
-   :lineno-start: 63
-   :lines: 27,64-136
    :linenos:
+   :pyobject: multiscale_icp
+   :lineno-match:
 
 Two options are given for the fine-grained registration. The ``color`` option is
 recommended since it uses color information to prevent drift. See [Park2017]_
@@ -33,9 +33,9 @@ Multiway registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/refine_registration.py
    :language: python
-   :lineno-start: 40
-   :lines: 27,41-63
    :linenos:
+   :pyobject: update_posegraph_for_scene
+   :lineno-match:
 
 This script uses the technique demonstrated in :ref:`/tutorial/pipelines/multiway_registration.ipynb`. Function ``update_posegraph_for_scene`` builds a pose graph for multiway registration of all fragments. Each graph node represents a fragment and its pose which transforms the geometry to the global space.
 
@@ -44,21 +44,20 @@ for multiway registration.
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/optimize_posegraph.py
    :language: python
-   :lineno-start: 63
-   :lines: 27,64-73
    :linenos:
+   :pyobject: optimize_posegraph_for_scene
+   :lineno-match:
 
 Main registration loop
 ``````````````````````````````````````
 
-The function ``make_posegraph_for_refined_scene`` below calls all the functions
- introduced above.
+The function ``make_posegraph_for_refined_scene`` below calls all the functions introduced above.
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/refine_registration.py
    :language: python
-   :lineno-start: 173
-   :lines: 27,174-223
    :linenos:
+   :pyobject: make_posegraph_for_refined_scene
+   :lineno-match:
 
 The main workflow is: pairwise local refinement -> multiway registration.
 

diff --git a/docs/tutorial/reconstruction_system/register_fragments.rst b/docs/tutorial/reconstruction_system/register_fragments.rst
@@ -20,9 +20,9 @@ Preprocess point cloud
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/register_fragments.py
    :language: python
-   :lineno-start: 41
-   :lines: 27,42-54
+   :pyobject: preprocess_point_cloud
    :linenos:
+   :lineno-match:
 
 This function downsamples a point cloud to make it sparser and regularly
 distributed. Normals and FPFH feature are precomputed. See
@@ -36,9 +36,9 @@ Compute initial registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/register_fragments.py
    :language: python
-   :lineno-start: 85
-   :lines: 27,86-114
+   :pyobject: compute_initial_registration
    :linenos:
+   :lineno-match:
 
 This function computes a rough alignment between two fragments. If the fragments
 are neighboring fragments, the rough alignment is determined by an aggregating
@@ -53,9 +53,9 @@ Pairwise global registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/register_fragments.py
    :language: python
-   :lineno-start: 54
-   :lines: 27,55-85
+   :pyobject: register_point_cloud_fpfh
    :linenos:
+   :lineno-match:
 
 This function uses :ref:`/tutorial/pipelines/global_registration.ipynb#RANSAC` or :ref:`/tutorial/pipelines/global_registration.ipynb#fast-global-registration` for pairwise global registration.
 
@@ -66,9 +66,9 @@ Multiway registration
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/register_fragments.py
    :language: python
-   :lineno-start: 114
-   :lines: 27,115-137
+   :pyobject: update_posegraph_for_scene
    :linenos:
+   :lineno-match:
 
 This script uses the technique demonstrated in
 :ref:`/tutorial/pipelines/multiway_registration.ipynb`. The function
@@ -81,9 +81,9 @@ called for multiway registration.
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/optimize_posegraph.py
    :language: python
-   :lineno-start: 63
-   :lines: 27,64-73
+   :pyobject: optimize_posegraph_for_scene
    :linenos:
+   :lineno-match:
 
 Main registration loop
 ``````````````````````````````````````
@@ -94,9 +94,9 @@ multiway registration.
 
 .. literalinclude:: ../../../examples/python/reconstruction_system/register_fragments.py
    :language: python
-   :lineno-start: 167
-   :lines: 27,168-210
+   :pyobject: make_posegraph_for_scene
    :linenos:
+   :lineno-match:
 
 Results
 ``````````````````````````````````````

diff --git a/docs/tutorial/reconstruction_system/system_overview.rst b/docs/tutorial/reconstruction_system/system_overview.rst
@@ -48,7 +48,7 @@ Getting the example code
 
 .. code-block:: sh
 
-    # Activate your conda enviroment, where you have installed open3d pip package.
+    # Activate your conda environment, where you have installed open3d pip package.
     # Clone the Open3D github repository and go to the example.
     cd examples/python/reconstruction_system/
 
@@ -69,7 +69,7 @@ Running the example with default dataset.
     python run_system.py --make --register --refine --integrate
 
 Changing the default dataset.
-One may change the default dataset to other avaialble datasets. 
+One may change the default dataset to other available datasets.
 Currently the following datasets are available:
 
 1. Lounge (keyword: ``lounge``) (Default)
@@ -81,7 +81,7 @@ Currently the following datasets are available:
 
 .. code-block:: sh
 
-    # Using jack_jack as the default dataset.
+    # Using bedroom as the default dataset.
     python run_system.py --default_dataset 'bedroom' --make --register --refine --integrate
 
 Running the example with custom dataset using config file.