r"""Average reduce a tensor across all workers.

r""" Compute matching matrix and score matrix for each patch correspondence. Input: - score_mat (Tensor): (B, K, K), or (B, K+

build instance-labeled points and instance list. Return: instance_list: (N+M,), instance_points: (P+Q, 3)

r"""Convert stacked correspondences to batched points. The extracted dense correspondences from all patch correspondences are stacked. Howeve

Create rotation matrices from a list of axes and angles. Code from wikipedia on quaternions :param axis: float32[N, 3] :param angle: floa

Read a batch of scene pairs. Encode all of their fine points and features.

background (floors, carpets) are considered. - gt_matrix: (n,m) - pred: (a,2), [i,j] where i in [0,n), j in [0,m) - return: true_pos, fals

find gt association Return: - matche_results: [(src_idx, tar_idx, iou)] - correspondences: np.array, (Mp,4), [src_idx, tar_idx,

Frees all allocated memory chunks */

Generate edge indices for a random graph with N nodes, where each node is connected to K random nodes (excluding itself). Input: - N:

r"""Compose transformation matrix from rotation matrix and translation vector. Args: rotation (Tensor): (*, 3, 3) translation (Te

Global average pooling over batch. Args: x: [N, D] input features batch_lengths: [B] list of batch lengths Returns:

Grid subsampling in stack mode. This function is implemented on CPU. Args: points (Tensor): stacked points. (N, 3) lengths (

Initialize the kernel point positions in a sphere.

r"""Compute the isotropic Relative Rotation Error and Relative Translation Error. Args: gt_transforms (Tensor): ground truth transformati

Creation of kernel point via optimization of potentials. :param radius: Radius of the kernels :param num_points: points composing kernels

r"""K-NN interpolate. WARNING: this function assumes the neighbors are ordered. Args: s_feats (Tensor): (M, C) q_points (Ten

instance_matches: (num_matches,3), [src_id,tar_id,iou] ious_map: dict, {'iou':iou_mat, 'src_names':src_names

r"""Load weights and check keys.

Args: ref_knn_points (Tensor): (B, K, 3), (128, 64, 3) src_knn_points (Tensor): (B, K, 3) score_mat (Tens

* Returns a pointer to a piece of new memory of the given size in bytes * allocated from the pool. */

Graph match semantic nodes. Return dict: - output_dict[pred_nodes]: (M,3),[batch_id,src_idx,tar_idx],

Input, - emb_indices: (N',K)

Query for the \a num_closest closest points to a given point (entered as * query_point[0:dim-1]). Note that this is a short-cut method for * ind

* Find all the neighbors to \a query_point[0:dim-1] within a maximum radius. * The output is given as a vector of pairs, of which the first elemen

r"""Isotropic Relative Rotation Error. RRE = acos((trace(R^T \cdot \bar{R}) - 1) / 2) Args: gt_rotations (Tensor): ground truth rota

r"""Isotropic Relative Rotation Error. RTE = \lVert t - \bar{t} \rVert_2 Args: gt_translations (Tensor): ground truth translation ve

Render the centers of nodes in the scene graph.

Render a point cloud with a specific color and point size.

desc0, desc1: (m,d) return: (m,n)

create the log assignment matrix from logits and similarity

Creation of kernel point via Lloyd algorithm. We use an approximation of the algorithm, and compute the Voronoi cell centers with discretizat

Split the instance_feats_dict into ref_dict and src_dict.

Compute the metrics: - nodes_tp, nodes_fp, nodes_gt - recall, precision, rmse, scenes - (optional) gt_precision

r"""Compute rigid transformation from `src_points` to `ref_points` using weighted SVD. Modified from [PointDSC](https://github.com/XuyangBai/Poin

finds position of least significant unset bit */

Generate by DeepSeekV3

Constructor: takes a const ref to the matrix object with the data points

Deleted copy constructor*/

* KDTree constructor * * Refer to docs in README.md or online in * https://github.com/jlblancoc/nanoflann * * The KD-Tree point dimensi

* KDTree constructor * * Refer to docs in README.md or online in * https://github.com/jlblancoc/nanoflann * * The KD-Tree point dimensi

Default constructor. Initializes a new pool. */

Note: The first argument (checks_IGNORED_) is ignored, but kept for * compatibility with the FLANN interface */