experimental_data_creation
Class FeatureFilteringIntrinsicData
java.lang.Object
experimental_data_creation.FeatureFilteringIntrinsicData
public class FeatureFilteringIntrinsicData
- extends java.lang.Object
This class can be used to generate the datasets of feature filtering based on intrinsic feature structure
experiment.
Perhaps we could learn new codebooks and new pca matrices using the only the retained features for each
image.
- Author:
- Eleftherios Spyromitros-Xioufis
Method Summary |
static void |
main(java.lang.String[] args)
|
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
numCentroids
public static final int numCentroids
- The number of centroids in the supplied codebooks.
- See Also:
- Constant Field Values
descriptorLength
public static final int descriptorLength
- See Also:
- Constant Field Values
FeatureFilteringIntrinsicData
public FeatureFilteringIntrinsicData()
main
public static void main(java.lang.String[] args)
throws java.lang.Exception
- Parameters:
args
- [0] Full path to the folder that contains the raw SURF feature files in text or binary
format.args
- [1] Full path to the folder where the BDB store will be created.args
- [2] Full path to a codebook file with 64 centroids learned using L2-normalized SURF
features.args
- [3] Percentage of features to be retained (valid percentages are: 0.5, 0.8, 0.9, 0.95).args
- [4] Filtering method to be applied (random/entropy/variance).
Additional parameters used when the entropy-based filtering method is selected.
args
- [5] Number of equal-width bins to be used for discretization.args
- [6] Full path to the serialized discretization filter file.args
- [7] Full path to the serialized un-discretized Instances object.
- Throws:
java.lang.Exception