Todo List

Namespace vt

Include some basic_types.h instead

Member vt::Database::DocumentVector

Use sorted vector?

Member vt::Database::find (const std::vector< Word > &document, size_t N, std::vector< Match > &matches) const

Try only computing distances against documents sharing at least one word

Member vt::Database::findAndInsert (const std::vector< Word > &document, size_t N, std::vector< Match > &matches)

Can this be accelerated? Could iterate over words only once?

Member vt::Database::insert (const std::vector< Word > &document)

Evaluate whether sorting words makes much difference in speed

Class vt::distance::L2< Eigen::Matrix< Scalar, Rows, Cols, Options, MaxRows, MaxCols > >

Version for raw data pointers that knows the size of the feature

Specialization for cv::Vec. Doesn't have size() so default won't work.

Member vt::GenericTree::save (const std::string &file) const

Currently assuming float. Really need more info in the save format.

Member vt::SimpleKmeans::clusterOnce (const std::vector< Feature * > &features, size_t k, std::vector< Feature, FeatureAllocator > &centers, std::vector< unsigned int > &membership) const

Kahan summation?

Member vt::SimpleKmeans::SimpleKmeans (const Feature &zero=Feature(), Distance d=Distance())

FeatureAllocator parameter

Class vt::SimpleKmeans< Feature, Distance, FeatureAllocator >

InitKmeanspp

Member vt::VocabularyTree::distance_

Consider bit-vector

Member vt::VocabularyTree::save (const std::string &file) const

Support serializing of non-"simple" feature classes

Some identifying name for the distance used

Member vt::VocabularyTree::VocabularyTree (Distance d=Distance())

Allocator parameter, also in MutableVocabularyTree, TreeBuilder...