Brute-force localization, as proposed in Fulkerson et. al 2008.
BK = BLOCK_TEST_BRUTELOC() Initializes the block with the default options.
BK = BLOCK_TEST_BRUTELOC(BK) Executes the block with options and inputs BK.
Required inputs:
- db
The database.
- feat
Features extracted on the database
- hist
Histograms extracted on the database.
- svm
A trained SVM.
- kernel
The kernel the SVM was trained with.
Options:
- bk.rand_seed
Default [] does not change the random seeds.
- bk.scaleby
Downsample the image by this factor before classification. Default 4.
- bk.windowsize
Half-width of the window to use in hisogram computation. Expressed in terms of the non-downsampled image. Default 80.
- bk.classifier
The type of classifier to use. Valid options: 'svm' or 'nn'. Default 'svm'.
- bk.seg_ids
The segment ids to classify. Default [] uses all testing images in the database.
Fetchable attributes:
- test
The classification result. Returns [class confidence] for required input: seg_id.