Creates an SE3 transform from translation and Euler angles. Args: xyzrpy (list[float]): translation and Euler angles for transform. Must

Args: code_weights: (#codes) float list if not None. Code-wise weights.

Note that out_channels = dv.

Args: radii: list of float, list of radii to group with nsamples: list of int, number of samples in each ball query

Project a pointcloud into a spherical projection, range image. Args: current_vertex: raw point clouds Returns: proj_rang

:param mlp: list of int, spec of the pointnet before the global max_pool :param npoint: int, number of features :param radius

:param pts: (N, 3 or 2) :return pts_hom: (N, 4 or 3)

Converts Euler angles to an SO3 rotation matrix. Args: rpy (list[float]): Euler angles (in radians). Must have three components. Ret

Converts Euler angles to an SO3 rotation matrix. Args: rpy (list[float]): Euler angles (in radians). Must have three components. Ret

Args: num_clusters : int The number of clusters dim : int Dimension of descriptors

Return an activation function given a string

Args: aggregate_func: xyz: (N, 3) xyz_features: (N, C) xyz_bs_idxs: (N) new_xyz:

Builds a pointcloud by combining multiple LIDAR scans with odometry information. Args: lidar_dir (str): Directory containing LIDAR scans.

Check that whether there are replicate indices in the sparse features, remove the replicate features if any.

Construct the multimodal features with both lidar sparse features and image features. Args: x: [N, C] lidar s

3x3 convolution with padding

Interpolate between absolute poses. Args: pose_timestamps (list[int]): Timestamps of supplied poses. Must be in ascending order.