Retrieval mAP vs. Vocabulary sizes
in Star Student Project Oxford Building data set and Star
Student Project Landmark dataset. Comparing methods in the above
subfigures include: Lossy Learning Star Student Project
Vocabulary Tree Star Student Project Tree Histogram Coding .
PCA based unsupervised BoW feature space compression. Note that
Approximate K Means on Oxford Buildings reported
in Star Student Project is also included for comparison. In
our Star Student Project Landmarks dataset, the
state-of-the-art compact vocabulary Star Student Project
scheme based on Tree Histogram Coding in Star Student Project
is also compared, as an alternative solution to What to distribute
Star Student Project . In addition, as can be seen from the
Star Student Project , PCA based selection is not as good as
our boosting based Star Student Project word selection.