org.openimaj.experiment.dataset.util

Class DatasetAdaptors

java.lang.Object

org.openimaj.experiment.dataset.util.DatasetAdaptors

public class DatasetAdaptors
extends Object

Helper methods to provide different types of view on a dataset.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
DatasetAdaptors()

Method Summary

Modifier and Type	Method and Description
static <INSTANCE> List<INSTANCE>	asList(Dataset<INSTANCE> dataset) Create a List view of the given dataset.
static <ANN,INSTANCE> GroupedDataset<ANN,ListDataset<INSTANCE>,INSTANCE>	flattenListGroupedDataset(GroupedDataset<ANN,? extends ListDataset<List<INSTANCE>>,? extends List<INSTANCE>> dataset) if you have a grouped dataset where the groups contains lists of feature objects (i.e.
static <ANN,DATASET extends Dataset<INSTANCE>,INSTANCE> GroupedDataset<ANN,DATASET,INSTANCE>	getGroupedDatasetSubset(GroupedDataset<ANN,DATASET,INSTANCE> data, ANN... groups) Takes a grouped dataset and returns a new dataset that contains only those groups specified.
static <ANN,DATASET extends ListDataset<INSTANCE>,INSTANCE> GroupedDataset<ANN,ListBackedDataset<INSTANCE>,INSTANCE>	getRegroupedDataset(GroupedDataset<ANN,DATASET,INSTANCE> data, Map<ANN,ANN[]> regroupCriteria) Takes a grouped dataset and returns a new dataset with the groups re-shuffled as specified in the regrouping criteria.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

DatasetAdaptors

public DatasetAdaptors()

Method Detail

asList

public static <INSTANCE> List<INSTANCE> asList(Dataset<INSTANCE> dataset)

Create a List view of the given dataset. If the dataset is a ListDataset it is returned, otherwise this method creates a new List containing all the instances in the dataset. The list is populated by iterating through the dataset.

Type Parameters:

INSTANCE - The type of instances in the dataset

Parameters:

dataset - The dataset.

Returns:

a list of all instances.

flattenListGroupedDataset

public static <ANN,INSTANCE> GroupedDataset<ANN,ListDataset<INSTANCE>,INSTANCE> flattenListGroupedDataset(GroupedDataset<ANN,? extends ListDataset<List<INSTANCE>>,? extends List<INSTANCE>> dataset)

if you have a grouped dataset where the groups contains lists of feature objects (i.e. GroupedDataset<KEY,ListDataset<List<INSTANCE>>,INSTANCE>) then this will flatten those internal list, so that all the instances from those lists are directly associated with the key. This type of thing might occur if your dataset element reader can extract multiple media parts from a single dataset item, that will all end up with the same key.

Parameters:

dataset - The dataset

Returns:

The new dataset

getGroupedDatasetSubset

@SafeVarargs
public static <ANN,DATASET extends Dataset<INSTANCE>,INSTANCE> GroupedDataset<ANN,DATASET,INSTANCE> getGroupedDatasetSubset(GroupedDataset<ANN,DATASET,INSTANCE> data,
                                                                                                                                         ANN... groups)

Takes a grouped dataset and returns a new dataset that contains only those groups specified. If the given groups do not exist in the provided dataset, then they will be ignored.

Parameters:

data - The dataset to take the groups from

groups - The groups to take

Returns:

the new dataset containing only those groups.

getRegroupedDataset

public static <ANN,DATASET extends ListDataset<INSTANCE>,INSTANCE> GroupedDataset<ANN,ListBackedDataset<INSTANCE>,INSTANCE> getRegroupedDataset(GroupedDataset<ANN,DATASET,INSTANCE> data,
                                                                                                                                                Map<ANN,ANN[]> regroupCriteria)

Takes a grouped dataset and returns a new dataset with the groups re-shuffled as specified in the regrouping criteria. The regrouping criteria is a map from new group name to old group name. Instances in the old group names will be mapped to the new group names. Where many old groups map to a single new group, the groups will be merged. For example:

 
        old == GroupedDataset: {G1=[1,2,3],G2=[4,5,6],G3=[7,8,9]}
 
        new = getGroupedDatasetSubset( old, {A->[G1,G3],B->[G2]} )
 
        new == GroupedDataset: {A=[1,2,3,7,8,9],B=[4,5,6]}
        
 

If the given groups do not exist in the provided dataset, then they will be ignored.

Parameters:

data - The dataset to take the groups from

regroupCriteria - The regrouping criteria

Returns:

the new dataset containing the new regrouping.