Generates an equivolumetric grid on SO(3) following Yershova et al. (2010). Uses a Healpix grid on the 2-sphere as a starting point and then tiles i

Returns the globally shared dictionary of data generation information, including success rate and trial number Returns:

Return the joint velocity(ies) of the arm. Args: joint_name (str, optional): If it's None, it will return

Translates meshgrid of xy pixel coordinates to normalized directions of rays through these pixels.

Translates meshgrid of uv pixel coordinates to normalized directions of rays through these pixels, in camera coordinates.

Function to check if a 3D point falls inside of the convex hull of a set of 3D points. A convex hull for the collision geometry is com

Generate relative transformation of bottle (child object) that is likely to be placed upright in a container (parent object), ideally

Generate relative transformation of bottle (child object) that is likely to be placed upright in a container (parent object), ideally

Generate relative transformation of bowl (child object) that is likely to be stacked on mug (parent object). Uses prior knowledge of

Generate relative transformation of can (child object) that is likely to be placed upright in a cabinet (parent object) in either an

Generate relative transformation of mug (child object) that is likely to hang on rack (parent object). Uses prior knowledge and heuri

Initilize the palm attributes based on some sampled palm poses. Also implements heuristic corrections of the palm samples based on the transfo

Setter function for internal surface attribute Args: skill_name (str): Name of skill that is used to reach this node

Initialize the pointcloud and transformation attributes based on some initial node (state) and sampled transformation. Args:

Setter function for internal surface attribute Args: surface_name (str): Name of surface that object is on at this node

Iteratively predict the relative transformation in one shot, applying each prediction to the shape and feeding the transformed shape back

Iteratively predict the relative transformation in one shot, applying each prediction to the shape and feeding the transformed shape back

Iteratively predict the relative transformation in one shot, applying each prediction to the shape and feeding the transformed shape back

Iteratively predict the relative transformation in one shot, applying each prediction to the shape and feeding the transformed shape back

Iteratively predict the relative transformation in one shot, applying each prediction to the shape and feeding the transformed shape back

from torch_cluster

:param self: :param x: Shape (batch_size, num_points) :param y: :param z: :param intrinsics: :return:

Loads config file. Args: path (str): path to config file default_path (bool): whether to use default path

Loads config file. Args: path (str): path to config file default_path (bool): whether to use default path

Converts rotation matrices to 6D rotation representation by Zhou et al. [1] by dropping the last row. Note that 6D representation is not uniq

Initializes coordinate axes of a frame T. The x-axis is drawn red, y-axis green and z-axis blue. The axes point in +x, +y and +z directions,

Move the end-effector in a straight line without changing the orientation. Args: delta_xyz (list or np.ndarray):

Relative pose predictions made in multiple iterative steps. Uses: - a final success classifier - potentially two versions of the re

Directly predict the relative transformation in one shot

Relative pose predictions made in multiple iterative steps. Uses: - a final success classifier - potentially two versions of the re

Normalize coordinate to [0, 1] for unit cube experiments. Corresponds to our 3D model Args: p (tensor): point padding (f

Normalize coordinate to [0, 1] for unit cube experiments. Corresponds to our 3D model Args: p (tensor): point padding (f

Helper function to initialize an open3d pointcloud object from a numpy array Args: points (np.ndarray): N X 3 pointcloud array