x: (b,n,d) encoding: (2,b,n,d) mask: (b,h,n,n)

x: (b,n,d) pos: (b,n,3) edge: (b,2,e) mask: (b,n,n)

Initialize nodes with instance features.

r"""KPConv forward. Args: s_feats (Tensor): (N, C_in) q_points (Tensor): (M, 3) s_points (Tensor): (N, 3)

Input: - matching_ptr: (M,) - ref_matching_points: (M,K,3) - src_matching_points: (M,K,3) - m

Generate attention mask between each pair of special tokens (eg. [CLS],[SEP]) Args: input_ids (torch.Tensor): input ids. Shape: [bs, num_t

Generate attention mask between each pair of special tokens Args: input_ids (torch.Tensor): input ids. Shape: [bs, num_token] spec

Creates multiple empty trees to handle dynamic support */

Make sure the auxiliary list \a vind has the same size than the current * dataset, and re-generate if size has changed. */

Input: - points_instances: (N, 1), fine points instances - instances: (M,), instance idx Output: - instance_f_points_

r"""Inverse rigid transform. Args: transform (Tensor): (*, 4, 4) Return: inv_transform (Tensor): (*, 4, 4)

check if the registration is successful source: (N,3), T_est: (4,4), T_gt: (4,4)

Returns the dim'th component of the idx'th point in the class: Since this is inlined and the "dim" argument is typically an immediate value, the "if/e

* Find the "num_closest" nearest neighbors to the \a query_point[0:dim-1]. * Their indices are stored inside the result object. \sa radiusSearch,

* Find the "num_closest" nearest neighbors to the \a query_point[0:dim-1]. * Their indices are stored inside the result object. \sa radiusSearch,

Loads a previous index from a binary file. * IMPORTANT NOTE: The set of data points is NOT stored in the file, so the * index object must be c

Loads a previous index from a binary file. * IMPORTANT NOTE: The set of data points is NOT stored in the file, so the * index object must be c

r"""Compute the modified chamfer distance. Args: raw_points (Tensor): (B, N_raw, 3) ref_points (Tensor): (B, N_ref, 3) sr

Assignment operator definiton */

* Subdivide the list of points by a plane perpendicular on axe corresponding * to the 'cutfeat' dimension at 'cutval' position. * * On ret

Sample point_limit points for each instance Input, - points: (N, 3), fine-level points - points_instances: (N, 1), fine point

* 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"""Collate function for registration in stack mode. Points are organized in the following order: [ref_1, ..., ref_B, src_1, ..., src_B]. The

r"""Release all tensors to item or numpy array.

Remove a point from the set (Lazy Deletion) */

r"""Reset seed for data loader worker.

r"""Compute the Rodrigues rotation matrix aligning source vectors to target vectors. Args: src_vectors (Tensor): source vectors (*, 3)

r"""Compute Rodrigues Rotation Matrix. R = I + \sin{\theta} K + (1 - \cos{\theta}) K^2, where K is the skew-symmetric matrix of the axis vect

From JiangLi's code https://github.com/llijiang/GuidedContrast/blob/main/model/semseg/semseg.py#L209

An instance contain invalid indices due to its number of points less than K. Fill the invalid indices by random sampling from the valid indic

An instance contain invalid indices due to its limited number of points. For each instance, sample K points from the valid indices. If th

Stores the index in a binary file. * IMPORTANT NOTE: The set of data points is NOT stored in the file, so when * loading the index object it m

Stores the index in a binary file. * IMPORTANT NOTE: The set of data points is NOT stored in the file, so when * loading the index object it m

r"""Saves the model checkpoint at a given epoch.

r"""Collate function for registration in stack mode. Points are organized in the following order: [ref_1, ..., ref_B, src_1, ..., src_B]. The

r"""Collate function for single point cloud in stack mode. Points are organized in the following order: [P_1, ..., P_B]. The correspondence i

Returns number of points in dataset */

* Computes the inde memory usage * Returns: memory used by the index */

Returns the length of each point in the dataset */

* Find the worst result (furtherest neighbor) without copying or sorting * Pre-conditions: size() > 0 */