vector_aggregation
Class VladAggregatorMultipleVocabularies
java.lang.Object
vector_aggregation.VladAggregatorMultipleVocabularies
public class VladAggregatorMultipleVocabularies
- extends java.lang.Object
This class computes multiple vocabulary VLAD vectors which are already power and L2 normalized.
- Author:
- Eleftherios Spyromitros-Xioufis
Method Summary |
double[] |
aggregate(java.util.ArrayList<double[]> descriptors)
Takes as input an ArrayList of double arrays which contains the set of local descriptors for an image. |
double[] |
aggregate(double[][] descriptors)
|
int |
getVectorLength()
|
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
vladAggregators
private VladAggregator[] vladAggregators
vectorLength
private int vectorLength
VladAggregatorMultipleVocabularies
public VladAggregatorMultipleVocabularies(double[][][] codebooks)
- The default constructor calls the super constructor.
- Parameters:
codebook
-
getVectorLength
public int getVectorLength()
aggregate
public double[] aggregate(java.util.ArrayList<double[]> descriptors)
throws java.lang.Exception
- Takes as input an ArrayList of double arrays which contains the set of local descriptors for an image.
Returns the VLAD vector representation of the image using the codebooks supplied in the constructor. TO
DO: A vectorized implementation would probably be faster.
- Parameters:
descriptors
-
- Returns:
-
- Throws:
java.lang.Exception
java.lang.Exception
aggregate
public double[] aggregate(double[][] descriptors)
throws java.lang.Exception
- Throws:
java.lang.Exception