orcanet.lib.sample_modifiers
Some basic sample modifiers to use with orcanet. Use them by setting .cfg.sample_modifier of the orcanet.core.Organizer.
Module Contents
Classes
For modifiers that do the same operation on each input. |
|
For applying multiple sample modifiers after each other. |
|
Permute the axes of the samples to given order. |
|
Reshape samples to given shape. |
|
Read out points and coordinates, intended for the MEdgeConv layers. |
- class orcanet.lib.sample_modifiers.PerInputModifier[source]
For modifiers that do the same operation on each input. Apply modify on x_value of each input, and output as dict.
- class orcanet.lib.sample_modifiers.JoinedModifier(sample_modifiers)[source]
For applying multiple sample modifiers after each other.
- class orcanet.lib.sample_modifiers.Permute(axes)[source]
Permute the axes of the samples to given order. Batchsize axis is excluded, i.e. start indexing with 1!
- class orcanet.lib.sample_modifiers.Reshape(newshape)[source]
Reshape samples to given shape. Batchsize axis is excluded!
- class orcanet.lib.sample_modifiers.GraphEdgeConv(knn=16, node_features=('pos_x', 'pos_y', 'pos_z', 'time', 'dir_x', 'dir_y', 'dir_z'), coord_features=('pos_x', 'pos_y', 'pos_z', 'time'), ragged=True, with_lightspeed=True, column_names=None, is_valid_features='is_valid', n_hits_padded=None)[source]
Read out points and coordinates, intended for the MEdgeConv layers.
For DL files produced with OrcaSong in graph mode.
- Parameters
- knnint or None
Number of nearest neighbors used in the edge conv. Pad events with too few hits by duping first hit, and give a warning.
- node_featurestuple
Defines the node features.
- coord_featurestuple
Defines the coordinates.
- raggedbool, optional
If True, return ragged tensors (nodes, coordinates). If False, return regular tensors, padded to fixed length. n_hits_padded and is_valid_features need to be given in this case.
- with_lightspeedbool
Multiply time for coordinates input with lightspeed. Requires coord_features to have the entry ‘time’.
- column_namestuple, optional
Name and order of the features in the last dimension of the array. If None is given, will attempt to auto-read the column names from the attributes of the dataset.
- is_valid_featuresstr
Only for when ragged = False. Defines the is_valid.
- n_hits_paddedint, optional
Only for when ragged = False. Pad or cut to exactly this many hits using 0s. Non-indexed datasets will automatically set this value.