cars_mesh.tools.metrics
=======================

.. py:module:: cars_mesh.tools.metrics

.. autoapi-nested-parse::

   Evaluation metrics



Classes
-------

.. autoapisummary::

   cars_mesh.tools.metrics.PointCloudMetrics


Functions
---------

.. autoapisummary::

   cars_mesh.tools.metrics.mean_squared_distance
   cars_mesh.tools.metrics.root_mean_squared_distance
   cars_mesh.tools.metrics.mean_distance
   cars_mesh.tools.metrics.median_distance
   cars_mesh.tools.metrics.hausdorff_asym_distance
   cars_mesh.tools.metrics.hausdorff_sym_distance
   cars_mesh.tools.metrics.chamfer_distance
   cars_mesh.tools.metrics.point_to_plane_distance


Module Contents
---------------

.. py:function:: mean_squared_distance(dist: numpy.array) -> float

.. py:function:: root_mean_squared_distance(dist: numpy.array) -> float

.. py:function:: mean_distance(dist: numpy.array) -> float

.. py:function:: median_distance(dist: numpy.array) -> float

.. py:function:: hausdorff_asym_distance(dist: numpy.array) -> float

.. py:function:: hausdorff_sym_distance(dist_1: numpy.array, dist_2: numpy.array) -> float

.. py:function:: chamfer_distance(dist_1: numpy.array, dist_2: numpy.array) -> float

.. py:function:: point_to_plane_distance(pcd_in, pcd_ref, knn=30, workers=1, use_open3d=True, **kwargs) -> numpy.ndarray

   The point-to-plane distance first computes the normal of the surface at
   every point in the reference point cloud as an indication of the local
   surface. The displacement of every corresponding point in the noisy
   point cloud is then projected onto the normal to calculate the
   point-to-plane distance.

   Source: Lang Zhou, Guoxing Sun, Yong Li, Weiqing Li, Zhiyong Su, Point
   cloud denoising review: from classical to deep learning-based
   approaches, Graphical Models, Volume 121, 2022, 101140, ISSN 1524-0703,
   https://doi.org/10.1016/j.gmod.2022.101140.


.. py:class:: PointCloudMetrics(pcd_in: cars_mesh.tools.handlers.PointCloud, pcd_ref: cars_mesh.tools.handlers.PointCloud, **kwargs)

   Compute metrics between two points clouds
   Nearest neighbours are computed during the initialisation step in order
   to share this information between different metrics and avoid a costly
   recomputing step.


   .. py:attribute:: pcd_in


   .. py:attribute:: pcd_ref


   .. py:attribute:: modes
      :value: ['p2p', 'p2s']



   .. py:attribute:: metrics


   .. py:attribute:: dist_p2p_in_ref


   .. py:attribute:: dist_p2p_ref_in


   .. py:attribute:: dist_p2s_in_ref


   .. py:attribute:: dist_p2s_ref_in


   .. py:method:: mean_squared_distance(mode) -> tuple

      Mean squared distance (or error, MSE)

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: root_mean_squared_distance(mode) -> tuple

      Root mean squared distance (or error, RMSE)

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: mean_distance(mode) -> tuple

      Mean distance (or error)

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: median_distance(mode) -> tuple

      Median distance (or error)

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: hausdorff_asym_distance(mode) -> tuple

      Hausdorff asymmetric distance

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: hausdorff_sym_distance(mode) -> float

      Hausdorff symmetric distance

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: chamfer_distance(mode) -> float

      Chamfer distance

      :param mode: Mode of computation. Either 'p2p' (point to point) or 'p2s'
                   (point to surface):
      :type mode: str



   .. py:method:: _serialize_ply_distances(filepath: str, pcd: cars_mesh.tools.handlers.PointCloud, distances: numpy.ndarray) -> None

      Serialize a textured mesh as a PLY file

      :param filepath: Filepath to the texture image
      :type filepath: str



   .. py:method:: visualize_distances(output_dir) -> None

      Save distances as a scalar field to be visualised on a viewer

      :param output_dir: Path to the output directory
      :type output_dir: str