Merge pull request #91 from autonomousvision/dev_v0.0.9

Merge `dev_v0.0.9` branch to `main`

Author: DanielDauner

.github/workflows/pytest.yaml

@@ -2,7 +2,7 @@ name: pytest
 
 on:
   push:
-    branches: [main]
+  pull_request:
 
 jobs:
   build:

.gitignore

@@ -26,6 +26,10 @@
 exp/
 .coverage
 
+# internal
+benchmarks/*
+notebooks/*
+
 # Sphinx documentation
 docs/_build/
 docs/build/
@@ -35,3 +39,6 @@ _build/
 
 # ruff
 .ruff_cache/*
+
+# Other
+CLAUDE.md

README.md

@@ -25,6 +25,18 @@
 
 ## Changelog
 
+- **`[2026-02-23]`** v0.0.9
+  - Added preliminary Waymo Open Motion Dataset support.
+  - Added support for nuScenes interpolated to 10Hz.
+  - Replaced gpkg map implementation with Arrow-based format for improved performance.
+  - Added sensor names and timestamps to camera and Lidar data across all datasets.
+  - Added ego-to-camera transforms in static metadata.
+  - Added support for loading merged point clouds in API.
+  - Improvements to geometry module, in terms of speed and syntax.
+  - Improved map querying speed and OpenDrive lane connectivity handling.
+  - Added recommended conversion options to dataset YAML configuration files.
+  - Improvements in dataset conversion for sensor sync, speed, and memory requirements.
+
 - **`[2025-11-21]`** v0.0.8 (silent release)
   - Release of package and documentation.
   - Demo data for tutorials.

docs/api/geometry/01_primitives/07_indexing_enums.rst

@@ -46,6 +46,27 @@ Poses
    :no-inherited-members:
 
 
+Matrices
+--------
+
+
+.. autoclass:: py123d.geometry.MatrixSO2Index
+   :members:
+   :no-inherited-members:
+
+.. autoclass:: py123d.geometry.MatrixSO3Index
+   :members:
+   :no-inherited-members:
+
+.. autoclass:: py123d.geometry.MatrixSE2Index
+   :members:
+   :no-inherited-members:
+
+.. autoclass:: py123d.geometry.MatrixSE3Index
+   :members:
+   :no-inherited-members:
+
+
 Bounding Boxes
 --------------
 

docs/api/geometry/02_transform/01_transform_2d.rst

@@ -1,30 +1,73 @@
 Transforms in 2D
 ^^^^^^^^^^^^^^^^
 
-Transform 2D points between frames
-----------------------------------
+Functions for converting SE(2) poses ``(x, y, yaw)`` and 2D points ``(x, y)``
+between coordinate frames. For the 3D counterpart see :doc:`02_transform_3d`.
 
-.. autofunction:: py123d.geometry.transform.convert_absolute_to_relative_points_2d_array
 
-.. autofunction:: py123d.geometry.transform.convert_relative_to_absolute_points_2d_array
+Convert SE2 poses between frames
+---------------------------------
 
-.. autofunction:: py123d.geometry.transform.convert_points_2d_array_between_origins
+Convert between absolute and relative coordinates:
 
+- :func:`~py123d.geometry.transform.abs_to_rel_se2_array` /
+  :func:`~py123d.geometry.transform.abs_to_rel_se2` --
+  absolute → relative: :math:`T_\text{rel} = T_\text{origin}^{-1} \cdot T_\text{abs}`
 
+- :func:`~py123d.geometry.transform.rel_to_abs_se2_array` /
+  :func:`~py123d.geometry.transform.rel_to_abs_se2` --
+  relative → absolute: :math:`T_\text{abs} = T_\text{origin} \cdot T_\text{rel}`
+
+- :func:`~py123d.geometry.transform.reframe_se2_array` /
+  :func:`~py123d.geometry.transform.reframe_se2` --
+  re-express poses from one reference frame to another.
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_se2_array
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_se2
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_se2_array
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_se2
+
+.. autofunction:: py123d.geometry.transform.reframe_se2_array
+
+.. autofunction:: py123d.geometry.transform.reframe_se2
 
-Transform SE2 poses between frames
-----------------------------------
 
-.. autofunction:: py123d.geometry.transform.convert_absolute_to_relative_se2_array
+Convert 2D points between frames
+---------------------------------
 
-.. autofunction:: py123d.geometry.transform.convert_relative_to_absolute_se2_array
+The same absolute/relative/reframe operations, applied to 2D points instead of
+SE2 poses:
 
-.. autofunction:: py123d.geometry.transform.convert_se2_array_between_origins
+- :func:`~py123d.geometry.transform.abs_to_rel_points_2d_array` /
+  :func:`~py123d.geometry.transform.abs_to_rel_point_2d`
 
+- :func:`~py123d.geometry.transform.rel_to_abs_points_2d_array` /
+  :func:`~py123d.geometry.transform.rel_to_abs_point_2d`
 
+- :func:`~py123d.geometry.transform.reframe_points_2d_array` /
+  :func:`~py123d.geometry.transform.reframe_point_2d`
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_points_2d_array
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_point_2d
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_points_2d_array
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_point_2d
+
+.. autofunction:: py123d.geometry.transform.reframe_points_2d_array
+
+.. autofunction:: py123d.geometry.transform.reframe_point_2d
+
+
+Translation along body-frame axes
+----------------------------------
 
-Translation along frame axes
-----------------------------
+Translate SE2 poses or 2D points along their local (body) coordinate axes.
+The yaw / orientation is preserved.
 
 .. autofunction:: py123d.geometry.transform.translate_se2_along_body_frame
 

docs/api/geometry/02_transform/02_transform_3d.rst

@@ -1,30 +1,73 @@
 Transforms in 3D
 ^^^^^^^^^^^^^^^^
 
-Transform 3D points between frames
-----------------------------------
+Functions for converting SE(3) poses ``(x, y, z, qw, qx, qy, qz)`` and 3D points
+``(x, y, z)`` between coordinate frames. Rotations are represented as unit
+quaternions. For the 2D counterpart see :doc:`01_transform_2d`.
 
-.. autofunction:: py123d.geometry.transform.convert_absolute_to_relative_points_3d_array
 
-.. autofunction:: py123d.geometry.transform.convert_relative_to_absolute_points_3d_array
+Convert SE3 poses between frames
+---------------------------------
 
-.. autofunction:: py123d.geometry.transform.convert_points_3d_array_between_origins
+Convert between absolute and relative coordinates:
 
+- :func:`~py123d.geometry.transform.abs_to_rel_se3_array` /
+  :func:`~py123d.geometry.transform.abs_to_rel_se3` --
+  absolute → relative: :math:`T_\text{rel} = T_\text{origin}^{-1} \cdot T_\text{abs}`
 
+- :func:`~py123d.geometry.transform.rel_to_abs_se3_array` /
+  :func:`~py123d.geometry.transform.rel_to_abs_se3` --
+  relative → absolute: :math:`T_\text{abs} = T_\text{origin} \cdot T_\text{rel}`
 
-Transform SE3 poses between frames
-----------------------------------
+- :func:`~py123d.geometry.transform.reframe_se3_array` /
+  :func:`~py123d.geometry.transform.reframe_se3` --
+  re-express poses from one reference frame to another.
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_se3_array
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_se3
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_se3_array
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_se3
+
+.. autofunction:: py123d.geometry.transform.reframe_se3_array
+
+.. autofunction:: py123d.geometry.transform.reframe_se3
+
+Convert 3D points between frames
+---------------------------------
 
-.. autofunction:: py123d.geometry.transform.convert_absolute_to_relative_se3_array
+The same absolute/relative/reframe operations, applied to 3D points instead of
+SE3 poses:
 
-.. autofunction:: py123d.geometry.transform.convert_relative_to_absolute_se3_array
+- :func:`~py123d.geometry.transform.abs_to_rel_points_3d_array` /
+  :func:`~py123d.geometry.transform.abs_to_rel_point_3d`
 
-.. autofunction:: py123d.geometry.transform.convert_se3_array_between_origins
+- :func:`~py123d.geometry.transform.rel_to_abs_points_3d_array` /
+  :func:`~py123d.geometry.transform.rel_to_abs_point_3d`
 
+- :func:`~py123d.geometry.transform.reframe_points_3d_array` /
+  :func:`~py123d.geometry.transform.reframe_point_3d`
 
+.. autofunction:: py123d.geometry.transform.abs_to_rel_points_3d_array
+
+.. autofunction:: py123d.geometry.transform.abs_to_rel_point_3d
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_points_3d_array
+
+.. autofunction:: py123d.geometry.transform.rel_to_abs_point_3d
+
+.. autofunction:: py123d.geometry.transform.reframe_points_3d_array
+
+.. autofunction:: py123d.geometry.transform.reframe_point_3d
+
+
+Translation along body-frame axes
+----------------------------------
 
-Translation along frame axes
-----------------------------
+Translate SE3 poses or 3D points along their local (body) coordinate axes.
+The orientation is preserved.
 
 .. autofunction:: py123d.geometry.transform.translate_se3_along_body_frame
 

docs/api/geometry/02_transform/index.rst

@@ -1,10 +1,19 @@
 Transforms
 ----------
 
+Coordinate-frame transformations for SE(2) and SE(3) poses and points.
+
+Each function comes in two variants:
+
+- **Array functions** (suffix ``_array``) operate on raw NumPy arrays and support
+  batch dimensions.
+- **Typed functions** (no suffix) accept and return typed geometry objects
+  (:class:`~py123d.geometry.PoseSE2`, :class:`~py123d.geometry.PoseSE3`,
+  :class:`~py123d.geometry.Point2D`, :class:`~py123d.geometry.Point3D`).
+
 
 .. toctree::
    :maxdepth: 2
 
-
    01_transform_2d
    02_transform_3d

docs/conf.py

@@ -10,7 +10,7 @@
 project = "py123d"
 copyright = "2025"
 author = "DanielDauner"
-release = "v0.0.8"
+release = "v0.0.9"
 
 # -- General configuration ---------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration

docs/datasets/av2.rst

@@ -162,26 +162,28 @@ You can also override the file path and run:
 
 .. code-block:: bash
 
-  py123d-conversion datasets=["av2_sensor_dataset"] \
+  py123d-conversion datasets=["av2-sensor"] \
   dataset_paths.av2_data_root=$AV2_DATA_ROOT # optional if env variable is set
 
 
+.. note::
+  The conversion of AV2 by default does not store sensor data in the logs, but only relative file paths.
+  To change this behavior, you need to adapt the ``av2-sensor.yaml`` converter configuration.
 
 Dataset Issues
 ~~~~~~~~~~~~~~
 
 - **Ego Vehicle:** The vehicle parameters are partially estimated and may be subject to inaccuracies.
 
 
-
 Citation
 ~~~~~~~~
 
 If you use this dataset in your research, please cite:
 
 .. code-block:: bibtex
 
-  @article{Wilson2023NEURIPS,
+  @article{Wilson2021NEURIPS,
     author = {Benjamin Wilson and William Qi and Tanmay Agarwal and John Lambert and Jagjeet Singh and Siddhesh Khandelwal and Bowen Pan and Ratnesh Kumar and Andrew Hartnett and Jhony Kaesemodel Pontes and Deva Ramanan and Peter Carr and James Hays},
     title = {Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting},
     booktitle = {Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks (NeurIPS Datasets and Benchmarks 2021)},

docs/datasets/carla.rst

@@ -7,6 +7,9 @@ CARLA is an open-source simulator for autonomous driving research.
 As such CARLA data is synthetic and can be generated with varying sensor and environmental conditions.
 The following documentation is largely incomplete and merely describes the provided demo data.
 
+.. note::
+  Data from the CARLA simulator can be collected using the `LEAD framework <https://github.com/autonomousvision/lead>`_, which provides a state-of-the-art expert driver in CARLA.
+
 .. dropdown:: Quick Links
   :open: