Debug Friendly assertion Allows to put a breakpoint, and catch any assertion error in debug

Checks the size of a tensor along all its dimensions, against a list of sizes The tensor must have the same number of dimensions as the list

Returns the Least Square Weighting scheme parameters

Returns a pose matrix tensor from a pose parameters tensor Parameters ---------- params_tensor : torch.Tensor

Close the metrics file

Loads the configuration of the filter

Parameters ---------- pmap: torch.Tensor [..., num_channels, H, W] A projection map tensor with : the last two dimens

Clears the current data stored in the backend

Returns ------- int : The number of parameters (rotation + translation) for this representation

Builds a projection image from a PointCloud (torch.Tensor) Parameters ---------- pointcloud : torch.Tensor

Returns the vertex index in the pose graph of the param block given a pose index

Applies a transformation to a point cloud Parameters ---------- points_3d : [B, N, 3] A X, Y, Z point cl

Returns the tensor of the parameters of the pose Parameters ---------- pose_matrix_tensor : torch.Tensor

Parameters ---------- tensor : torch.Tensor A tensor which to be filtered for null points. null points must be null across

Returns the key in the keyframe dict for a new pose

Returns the Point-to-Plane residual closure The returned closure takes input a pose matrix or pose params tensor, And return

Returns the key in the keyframe dict for a new pointcloud

Removes all `nan` values from a one or two dimensional tensor

Returns the Attenuation Factor used to define the weighted least square

The column delimiter in pandas csv

An initialization procedure called at the start of each sequence

Returns the key where the initial pose estimate is saved

Returns ------- int : The number of parameters of rotation for this representation

Reads an array of poses from disk Returns ---------- poses : np.ndarray [N, 4, 4]

Updates the Local Map, by incorporating the new frame registered Args: new_relative_pose (torch.Tensor): The relative po

Computes voxel hashes for an array of voxel indices Args: voxel (np.ndarray): An array of voxels `(n, 3)` [np.int64] see https:

Computes voxel coordinates for a given pointcloud Args: pointcloud (np.ndarray): The input pointcloud `(n, 3)` [np.float32]

The key (in data_dict) for the absolute_pose_gt

Returns the Point-to-Plane residual jacobian closure The returned closure takes input a pose matrix or pose params tensor, A

Converts an array of pose matrices [N, 4, 4] to a DataFrame [N, 12]

Returns the prediction module for the specific trainer Returns ------- nn.Module The evaluation module w

Saves the metrics dictionary saved as a yaml file

Returns a key which defines a relative constraint between two consecutive poses Backends will search the `dict` of a new frame for keys follo

Returns the train, eval and test datasets and the corresponding sequence name Returns: (train, eval, test, transform) tr

Launches num_epochs of training Parameters ---------- num_epochs : int The number of epochs to launch

Applies the transform `tr` to the pointcloud Parameters ---------- pointcloud : np.ndarray (N, 3) tr : np.ndarray (4, 4)

Returns the poses expressed in the first pose frame Note: Should differ from world_poses only when additional information on the trajectory a

Attempts to align the target onto the reference, and if enough inliers are found, If succesful, returns the planar transform

Builds the KdTree and initialize the computation of the normals

Estimates the optimal set of parameters for a LeastSquare problem Args: x0 (torch.Tensor): The initial set of parameters

Projects the pc into the image plane, and compute features and descriptors

Computes the nearest neighbors between a target vertex map and a batch of reference vertex maps Args: vm_target (torch.Tensor): The

Computes the normals for a VertexMap (An image of X,Y,Z points projected into the image plane) Parameters ---------- vertex_map

Corrects the calibration of KITTI's HDL-64 scan

Draws multiple 2D trajectories in matplotlib plots and saves the plots as png images Parameters ---------- xs : list of arrays

Returns the euler convention used for the parametrisation of the rotation Fails if self.pose_type is not equal to "euler"

Safely retrieves the hash of the last commit If any failure, returns None

Returns the poses of the sequence of frames started after init() was called

Returns the poses of a given drive

Returns a color image from a gray image Parameters: np_image (np.ndarray): a gray image `(H,W)` color_map (str): The name of

Returns the ground truth associated with the sequence

Handles Failure cases of the SLAM runner

Clears the current representation and prepare a new model

Reads and Returns a Lidar Scan from kitti's benchmark Args: file_path (str): The file path of kitti's binary file Returns:

Converts a LLU GPS position to ECEF frame

Loads the Datasets for which the odometry is evaluated Returns ------- A list of pairs (sequence_name :str, dataset_

Returns the SLAM algorithm which will be run

Logs the output of the Parameters ---------- log_dict : dict The dictionary of items to be logged

Processes a next frame