print images id which can see the points :param points_seen: (N, M), seen flag of all points in all views

Runs a grid search on the partition parameters to find the best setup on a given sample `dataset[i_cloud]`. :param model: PanopticSegmentatio

Whether the edges have features in `edge_attr`.

Whether the points have edges.

Check whether `self` contains predicted data for panoptic segmentation `obj_index_pred`.

Check whether `self` contains predicted data for panoptic segmentation `obj_index_pred` as a list of results for performance compariso

Whether the points have neighbors.

Check whether `self` contains target data for panoptic segmentation.

Check whether `self` contains target data for semantic segmentation.

Convert ground truth and predictions at a segment level (i.e. ground truth is 2D tensor carrying histogram of labels in each segment), to poin

Read the host machine's name and return the known $DATA_ROOT directory

Returns the dense indices corresponding to the pointers.

Manual weight initialization. Allows setting specific init modes for certain modules. In particular, the linear and RPE layers are initialized

Compute the oracle performance for instance segmentation. This is a proxy for the highest achievable performance with the cluster part

Instantiates callbacks from config.

Instantiates loggers from config.

根据相机内参计算视场角。 参数: f_x (float): x轴的焦距（像素单位） f_y (float): y轴的焦距（像素单位） width (int): 图像宽度（像素单位） height (int): 图像高度（像素单位） 返回:

Return True is one of the Data contains a CUDA Tensor.

Key name for is_index_value. This will be used as labels for serialization.

Whether the points belong to a coarser super-graph.

Whether the points are superpoints for a denser sub-graph.

Ratios of number of nodes between consecutive partition levels. This can be useful for investigating how much each partition level 'co

Decorating a function will log how long it took to execute each function call Usage: @linetimer() def foo(): pass "Cod

listify `arg_ref` and the `args`, while ensuring that the length of `args` match the length of `arg_ref`. This is typically needed for parsing

Load ClusterBatch from an HDF5 file. See `Cluster.save` for writing such file. Options allow reading only part of the clusters.

Read an HDF5 file and return its content as a Batch object. See Data.load()

Load CSRBatch from an HDF5 file. See `CSRData.save` for writing such file. Options allow reading only part of the clusters. :

Controls which config parts are saved by lightning loggers. Additionally saves: - Number of model parameters

Generate an ipywidget for locally browsing a checkpoint file

Generate an ipywidget for selecting the device on which to work

Generate two co-dependent ipywidgets for selecting the task and experiment from a predefined set of experiment configs.

Generate an ipywidget for selecting the data split on which to work

If the semantic segmentation `logits` are stored in an enumerable, then the model output is multi-stage.

:param seg_seen: (S, M), ratio of seen part of primitives in every view :param seg_neighbors: (S, S), 1 if two segs are neighbors :param

Analyzes the key in a natural way to sort the strings that contain numbers. It converts the string into a list of mixed types

Index to be used for LayerNorm. :param mode: str Normalization mode. 'graph' will normalize per graph (i.e. per cloud

Number for semantic classes in the output predictions.

Number for edges in the instance graph.

Overwrite the torch_geometric initial definition, which somehow returns incorrect results, like: data.num_edges != data.edge_index

Number of levels of hierarchy in the nested graph.

Number for nodes/superpoints in the output predictions. By default, for a hierarchical partition, this means counting the number of le

Number of points/nodes in the lower-level graph.

This assumes all items in `self.obj_index_pred` follow the output format of `InstancePartitioner._grid_forward()`.

InstanceData object indicating the instance indices for each node/point/superpoint in the Data.

Compute an oracle for superpoint clustering for instance and panoptic segmentation. This is a proxy for the highest achievable graph clusterin

Return the predicted InstanceData, and the predicted instance semantic label and score, for a given batch item and a given partition s

Compute the oracle performance for panoptic segmentation. This is a proxy for the highest achievable performance with the cluster part

Key name for pointers. This will be used as labels for serialization.