Renesas / opencv-lib

openCV library for Renesas RZ/A

Dependents:   RZ_A2M_Mbed_samples

include/opencv2/flann/lsh_index.h@0:0e0631af0305, 2021-01-29 (annotated)

Committer:
RyoheiHagimoto
Date:
Fri Jan 29 04:53:38 2021 +0000
Revision:
0:0e0631af0305
copied from https://github.com/d-kato/opencv-lib.

Who changed what in which revision?

UserRevisionLine numberNew contents of line
RyoheiHagimoto 0:0e0631af0305 1 /***********************************************************************
RyoheiHagimoto 0:0e0631af0305 2 * Software License Agreement (BSD License)
RyoheiHagimoto 0:0e0631af0305 3 *
RyoheiHagimoto 0:0e0631af0305 4 * Copyright 2008-2009 Marius Muja (mariusm@cs.ubc.ca). All rights reserved.
RyoheiHagimoto 0:0e0631af0305 5 * Copyright 2008-2009 David G. Lowe (lowe@cs.ubc.ca). All rights reserved.
RyoheiHagimoto 0:0e0631af0305 6 *
RyoheiHagimoto 0:0e0631af0305 7 * THE BSD LICENSE
RyoheiHagimoto 0:0e0631af0305 8 *
RyoheiHagimoto 0:0e0631af0305 9 * Redistribution and use in source and binary forms, with or without
RyoheiHagimoto 0:0e0631af0305 10 * modification, are permitted provided that the following conditions
RyoheiHagimoto 0:0e0631af0305 11 * are met:
RyoheiHagimoto 0:0e0631af0305 12 *
RyoheiHagimoto 0:0e0631af0305 13 * 1. Redistributions of source code must retain the above copyright
RyoheiHagimoto 0:0e0631af0305 14 * notice, this list of conditions and the following disclaimer.
RyoheiHagimoto 0:0e0631af0305 15 * 2. Redistributions in binary form must reproduce the above copyright
RyoheiHagimoto 0:0e0631af0305 16 * notice, this list of conditions and the following disclaimer in the
RyoheiHagimoto 0:0e0631af0305 17 * documentation and/or other materials provided with the distribution.
RyoheiHagimoto 0:0e0631af0305 18 *
RyoheiHagimoto 0:0e0631af0305 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
RyoheiHagimoto 0:0e0631af0305 20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
RyoheiHagimoto 0:0e0631af0305 21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
RyoheiHagimoto 0:0e0631af0305 22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
RyoheiHagimoto 0:0e0631af0305 23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
RyoheiHagimoto 0:0e0631af0305 24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
RyoheiHagimoto 0:0e0631af0305 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
RyoheiHagimoto 0:0e0631af0305 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
RyoheiHagimoto 0:0e0631af0305 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
RyoheiHagimoto 0:0e0631af0305 28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
RyoheiHagimoto 0:0e0631af0305 29 *************************************************************************/
RyoheiHagimoto 0:0e0631af0305 30
RyoheiHagimoto 0:0e0631af0305 31 /***********************************************************************
RyoheiHagimoto 0:0e0631af0305 32 * Author: Vincent Rabaud
RyoheiHagimoto 0:0e0631af0305 33 *************************************************************************/
RyoheiHagimoto 0:0e0631af0305 34
RyoheiHagimoto 0:0e0631af0305 35 #ifndef OPENCV_FLANN_LSH_INDEX_H_
RyoheiHagimoto 0:0e0631af0305 36 #define OPENCV_FLANN_LSH_INDEX_H_
RyoheiHagimoto 0:0e0631af0305 37
RyoheiHagimoto 0:0e0631af0305 38 #include <algorithm>
RyoheiHagimoto 0:0e0631af0305 39 #include <cassert>
RyoheiHagimoto 0:0e0631af0305 40 #include <cstring>
RyoheiHagimoto 0:0e0631af0305 41 #include <map>
RyoheiHagimoto 0:0e0631af0305 42 #include <vector>
RyoheiHagimoto 0:0e0631af0305 43
RyoheiHagimoto 0:0e0631af0305 44 #include "general.h"
RyoheiHagimoto 0:0e0631af0305 45 #include "nn_index.h"
RyoheiHagimoto 0:0e0631af0305 46 #include "matrix.h"
RyoheiHagimoto 0:0e0631af0305 47 #include "result_set.h"
RyoheiHagimoto 0:0e0631af0305 48 #include "heap.h"
RyoheiHagimoto 0:0e0631af0305 49 #include "lsh_table.h"
RyoheiHagimoto 0:0e0631af0305 50 #include "allocator.h"
RyoheiHagimoto 0:0e0631af0305 51 #include "random.h"
RyoheiHagimoto 0:0e0631af0305 52 #include "saving.h"
RyoheiHagimoto 0:0e0631af0305 53
RyoheiHagimoto 0:0e0631af0305 54 namespace cvflann
RyoheiHagimoto 0:0e0631af0305 55 {
RyoheiHagimoto 0:0e0631af0305 56
RyoheiHagimoto 0:0e0631af0305 57 struct LshIndexParams : public IndexParams
RyoheiHagimoto 0:0e0631af0305 58 {
RyoheiHagimoto 0:0e0631af0305 59 LshIndexParams(unsigned int table_number = 12, unsigned int key_size = 20, unsigned int multi_probe_level = 2)
RyoheiHagimoto 0:0e0631af0305 60 {
RyoheiHagimoto 0:0e0631af0305 61 (* this)["algorithm"] = FLANN_INDEX_LSH;
RyoheiHagimoto 0:0e0631af0305 62 // The number of hash tables to use
RyoheiHagimoto 0:0e0631af0305 63 (*this)["table_number"] = table_number;
RyoheiHagimoto 0:0e0631af0305 64 // The length of the key in the hash tables
RyoheiHagimoto 0:0e0631af0305 65 (*this)["key_size"] = key_size;
RyoheiHagimoto 0:0e0631af0305 66 // Number of levels to use in multi-probe (0 for standard LSH)
RyoheiHagimoto 0:0e0631af0305 67 (*this)["multi_probe_level"] = multi_probe_level;
RyoheiHagimoto 0:0e0631af0305 68 }
RyoheiHagimoto 0:0e0631af0305 69 };
RyoheiHagimoto 0:0e0631af0305 70
RyoheiHagimoto 0:0e0631af0305 71 /**
RyoheiHagimoto 0:0e0631af0305 72 * Randomized kd-tree index
RyoheiHagimoto 0:0e0631af0305 73 *
RyoheiHagimoto 0:0e0631af0305 74 * Contains the k-d trees and other information for indexing a set of points
RyoheiHagimoto 0:0e0631af0305 75 * for nearest-neighbor matching.
RyoheiHagimoto 0:0e0631af0305 76 */
RyoheiHagimoto 0:0e0631af0305 77 template<typename Distance>
RyoheiHagimoto 0:0e0631af0305 78 class LshIndex : public NNIndex<Distance>
RyoheiHagimoto 0:0e0631af0305 79 {
RyoheiHagimoto 0:0e0631af0305 80 public:
RyoheiHagimoto 0:0e0631af0305 81 typedef typename Distance::ElementType ElementType;
RyoheiHagimoto 0:0e0631af0305 82 typedef typename Distance::ResultType DistanceType;
RyoheiHagimoto 0:0e0631af0305 83
RyoheiHagimoto 0:0e0631af0305 84 /** Constructor
RyoheiHagimoto 0:0e0631af0305 85 * @param input_data dataset with the input features
RyoheiHagimoto 0:0e0631af0305 86 * @param params parameters passed to the LSH algorithm
RyoheiHagimoto 0:0e0631af0305 87 * @param d the distance used
RyoheiHagimoto 0:0e0631af0305 88 */
RyoheiHagimoto 0:0e0631af0305 89 LshIndex(const Matrix<ElementType>& input_data, const IndexParams& params = LshIndexParams(),
RyoheiHagimoto 0:0e0631af0305 90 Distance d = Distance()) :
RyoheiHagimoto 0:0e0631af0305 91 dataset_(input_data), index_params_(params), distance_(d)
RyoheiHagimoto 0:0e0631af0305 92 {
RyoheiHagimoto 0:0e0631af0305 93 // cv::flann::IndexParams sets integer params as 'int', so it is used with get_param
RyoheiHagimoto 0:0e0631af0305 94 // in place of 'unsigned int'
RyoheiHagimoto 0:0e0631af0305 95 table_number_ = (unsigned int)get_param<int>(index_params_,"table_number",12);
RyoheiHagimoto 0:0e0631af0305 96 key_size_ = (unsigned int)get_param<int>(index_params_,"key_size",20);
RyoheiHagimoto 0:0e0631af0305 97 multi_probe_level_ = (unsigned int)get_param<int>(index_params_,"multi_probe_level",2);
RyoheiHagimoto 0:0e0631af0305 98
RyoheiHagimoto 0:0e0631af0305 99 feature_size_ = (unsigned)dataset_.cols;
RyoheiHagimoto 0:0e0631af0305 100 fill_xor_mask(0, key_size_, multi_probe_level_, xor_masks_);
RyoheiHagimoto 0:0e0631af0305 101 }
RyoheiHagimoto 0:0e0631af0305 102
RyoheiHagimoto 0:0e0631af0305 103
RyoheiHagimoto 0:0e0631af0305 104 LshIndex(const LshIndex&);
RyoheiHagimoto 0:0e0631af0305 105 LshIndex& operator=(const LshIndex&);
RyoheiHagimoto 0:0e0631af0305 106
RyoheiHagimoto 0:0e0631af0305 107 /**
RyoheiHagimoto 0:0e0631af0305 108 * Builds the index
RyoheiHagimoto 0:0e0631af0305 109 */
RyoheiHagimoto 0:0e0631af0305 110 void buildIndex()
RyoheiHagimoto 0:0e0631af0305 111 {
RyoheiHagimoto 0:0e0631af0305 112 tables_.resize(table_number_);
RyoheiHagimoto 0:0e0631af0305 113 for (unsigned int i = 0; i < table_number_; ++i) {
RyoheiHagimoto 0:0e0631af0305 114 lsh::LshTable<ElementType>& table = tables_[i];
RyoheiHagimoto 0:0e0631af0305 115 table = lsh::LshTable<ElementType>(feature_size_, key_size_);
RyoheiHagimoto 0:0e0631af0305 116
RyoheiHagimoto 0:0e0631af0305 117 // Add the features to the table
RyoheiHagimoto 0:0e0631af0305 118 table.add(dataset_);
RyoheiHagimoto 0:0e0631af0305 119 }
RyoheiHagimoto 0:0e0631af0305 120 }
RyoheiHagimoto 0:0e0631af0305 121
RyoheiHagimoto 0:0e0631af0305 122 flann_algorithm_t getType() const
RyoheiHagimoto 0:0e0631af0305 123 {
RyoheiHagimoto 0:0e0631af0305 124 return FLANN_INDEX_LSH;
RyoheiHagimoto 0:0e0631af0305 125 }
RyoheiHagimoto 0:0e0631af0305 126
RyoheiHagimoto 0:0e0631af0305 127
RyoheiHagimoto 0:0e0631af0305 128 void saveIndex(FILE* stream)
RyoheiHagimoto 0:0e0631af0305 129 {
RyoheiHagimoto 0:0e0631af0305 130 save_value(stream,table_number_);
RyoheiHagimoto 0:0e0631af0305 131 save_value(stream,key_size_);
RyoheiHagimoto 0:0e0631af0305 132 save_value(stream,multi_probe_level_);
RyoheiHagimoto 0:0e0631af0305 133 save_value(stream, dataset_);
RyoheiHagimoto 0:0e0631af0305 134 }
RyoheiHagimoto 0:0e0631af0305 135
RyoheiHagimoto 0:0e0631af0305 136 void loadIndex(FILE* stream)
RyoheiHagimoto 0:0e0631af0305 137 {
RyoheiHagimoto 0:0e0631af0305 138 load_value(stream, table_number_);
RyoheiHagimoto 0:0e0631af0305 139 load_value(stream, key_size_);
RyoheiHagimoto 0:0e0631af0305 140 load_value(stream, multi_probe_level_);
RyoheiHagimoto 0:0e0631af0305 141 load_value(stream, dataset_);
RyoheiHagimoto 0:0e0631af0305 142 // Building the index is so fast we can afford not storing it
RyoheiHagimoto 0:0e0631af0305 143 buildIndex();
RyoheiHagimoto 0:0e0631af0305 144
RyoheiHagimoto 0:0e0631af0305 145 index_params_["algorithm"] = getType();
RyoheiHagimoto 0:0e0631af0305 146 index_params_["table_number"] = table_number_;
RyoheiHagimoto 0:0e0631af0305 147 index_params_["key_size"] = key_size_;
RyoheiHagimoto 0:0e0631af0305 148 index_params_["multi_probe_level"] = multi_probe_level_;
RyoheiHagimoto 0:0e0631af0305 149 }
RyoheiHagimoto 0:0e0631af0305 150
RyoheiHagimoto 0:0e0631af0305 151 /**
RyoheiHagimoto 0:0e0631af0305 152 * Returns size of index.
RyoheiHagimoto 0:0e0631af0305 153 */
RyoheiHagimoto 0:0e0631af0305 154 size_t size() const
RyoheiHagimoto 0:0e0631af0305 155 {
RyoheiHagimoto 0:0e0631af0305 156 return dataset_.rows;
RyoheiHagimoto 0:0e0631af0305 157 }
RyoheiHagimoto 0:0e0631af0305 158
RyoheiHagimoto 0:0e0631af0305 159 /**
RyoheiHagimoto 0:0e0631af0305 160 * Returns the length of an index feature.
RyoheiHagimoto 0:0e0631af0305 161 */
RyoheiHagimoto 0:0e0631af0305 162 size_t veclen() const
RyoheiHagimoto 0:0e0631af0305 163 {
RyoheiHagimoto 0:0e0631af0305 164 return feature_size_;
RyoheiHagimoto 0:0e0631af0305 165 }
RyoheiHagimoto 0:0e0631af0305 166
RyoheiHagimoto 0:0e0631af0305 167 /**
RyoheiHagimoto 0:0e0631af0305 168 * Computes the index memory usage
RyoheiHagimoto 0:0e0631af0305 169 * Returns: memory used by the index
RyoheiHagimoto 0:0e0631af0305 170 */
RyoheiHagimoto 0:0e0631af0305 171 int usedMemory() const
RyoheiHagimoto 0:0e0631af0305 172 {
RyoheiHagimoto 0:0e0631af0305 173 return (int)(dataset_.rows * sizeof(int));
RyoheiHagimoto 0:0e0631af0305 174 }
RyoheiHagimoto 0:0e0631af0305 175
RyoheiHagimoto 0:0e0631af0305 176
RyoheiHagimoto 0:0e0631af0305 177 IndexParams getParameters() const
RyoheiHagimoto 0:0e0631af0305 178 {
RyoheiHagimoto 0:0e0631af0305 179 return index_params_;
RyoheiHagimoto 0:0e0631af0305 180 }
RyoheiHagimoto 0:0e0631af0305 181
RyoheiHagimoto 0:0e0631af0305 182 /**
RyoheiHagimoto 0:0e0631af0305 183 * \brief Perform k-nearest neighbor search
RyoheiHagimoto 0:0e0631af0305 184 * \param[in] queries The query points for which to find the nearest neighbors
RyoheiHagimoto 0:0e0631af0305 185 * \param[out] indices The indices of the nearest neighbors found
RyoheiHagimoto 0:0e0631af0305 186 * \param[out] dists Distances to the nearest neighbors found
RyoheiHagimoto 0:0e0631af0305 187 * \param[in] knn Number of nearest neighbors to return
RyoheiHagimoto 0:0e0631af0305 188 * \param[in] params Search parameters
RyoheiHagimoto 0:0e0631af0305 189 */
RyoheiHagimoto 0:0e0631af0305 190 virtual void knnSearch(const Matrix<ElementType>& queries, Matrix<int>& indices, Matrix<DistanceType>& dists, int knn, const SearchParams& params)
RyoheiHagimoto 0:0e0631af0305 191 {
RyoheiHagimoto 0:0e0631af0305 192 assert(queries.cols == veclen());
RyoheiHagimoto 0:0e0631af0305 193 assert(indices.rows >= queries.rows);
RyoheiHagimoto 0:0e0631af0305 194 assert(dists.rows >= queries.rows);
RyoheiHagimoto 0:0e0631af0305 195 assert(int(indices.cols) >= knn);
RyoheiHagimoto 0:0e0631af0305 196 assert(int(dists.cols) >= knn);
RyoheiHagimoto 0:0e0631af0305 197
RyoheiHagimoto 0:0e0631af0305 198
RyoheiHagimoto 0:0e0631af0305 199 KNNUniqueResultSet<DistanceType> resultSet(knn);
RyoheiHagimoto 0:0e0631af0305 200 for (size_t i = 0; i < queries.rows; i++) {
RyoheiHagimoto 0:0e0631af0305 201 resultSet.clear();
RyoheiHagimoto 0:0e0631af0305 202 std::fill_n(indices[i], knn, -1);
RyoheiHagimoto 0:0e0631af0305 203 std::fill_n(dists[i], knn, std::numeric_limits<DistanceType>::max());
RyoheiHagimoto 0:0e0631af0305 204 findNeighbors(resultSet, queries[i], params);
RyoheiHagimoto 0:0e0631af0305 205 if (get_param(params,"sorted",true)) resultSet.sortAndCopy(indices[i], dists[i], knn);
RyoheiHagimoto 0:0e0631af0305 206 else resultSet.copy(indices[i], dists[i], knn);
RyoheiHagimoto 0:0e0631af0305 207 }
RyoheiHagimoto 0:0e0631af0305 208 }
RyoheiHagimoto 0:0e0631af0305 209
RyoheiHagimoto 0:0e0631af0305 210
RyoheiHagimoto 0:0e0631af0305 211 /**
RyoheiHagimoto 0:0e0631af0305 212 * Find set of nearest neighbors to vec. Their indices are stored inside
RyoheiHagimoto 0:0e0631af0305 213 * the result object.
RyoheiHagimoto 0:0e0631af0305 214 *
RyoheiHagimoto 0:0e0631af0305 215 * Params:
RyoheiHagimoto 0:0e0631af0305 216 * result = the result object in which the indices of the nearest-neighbors are stored
RyoheiHagimoto 0:0e0631af0305 217 * vec = the vector for which to search the nearest neighbors
RyoheiHagimoto 0:0e0631af0305 218 * maxCheck = the maximum number of restarts (in a best-bin-first manner)
RyoheiHagimoto 0:0e0631af0305 219 */
RyoheiHagimoto 0:0e0631af0305 220 void findNeighbors(ResultSet<DistanceType>& result, const ElementType* vec, const SearchParams& /*searchParams*/)
RyoheiHagimoto 0:0e0631af0305 221 {
RyoheiHagimoto 0:0e0631af0305 222 getNeighbors(vec, result);
RyoheiHagimoto 0:0e0631af0305 223 }
RyoheiHagimoto 0:0e0631af0305 224
RyoheiHagimoto 0:0e0631af0305 225 private:
RyoheiHagimoto 0:0e0631af0305 226 /** Defines the comparator on score and index
RyoheiHagimoto 0:0e0631af0305 227 */
RyoheiHagimoto 0:0e0631af0305 228 typedef std::pair<float, unsigned int> ScoreIndexPair;
RyoheiHagimoto 0:0e0631af0305 229 struct SortScoreIndexPairOnSecond
RyoheiHagimoto 0:0e0631af0305 230 {
RyoheiHagimoto 0:0e0631af0305 231 bool operator()(const ScoreIndexPair& left, const ScoreIndexPair& right) const
RyoheiHagimoto 0:0e0631af0305 232 {
RyoheiHagimoto 0:0e0631af0305 233 return left.second < right.second;
RyoheiHagimoto 0:0e0631af0305 234 }
RyoheiHagimoto 0:0e0631af0305 235 };
RyoheiHagimoto 0:0e0631af0305 236
RyoheiHagimoto 0:0e0631af0305 237 /** Fills the different xor masks to use when getting the neighbors in multi-probe LSH
RyoheiHagimoto 0:0e0631af0305 238 * @param key the key we build neighbors from
RyoheiHagimoto 0:0e0631af0305 239 * @param lowest_index the lowest index of the bit set
RyoheiHagimoto 0:0e0631af0305 240 * @param level the multi-probe level we are at
RyoheiHagimoto 0:0e0631af0305 241 * @param xor_masks all the xor mask
RyoheiHagimoto 0:0e0631af0305 242 */
RyoheiHagimoto 0:0e0631af0305 243 void fill_xor_mask(lsh::BucketKey key, int lowest_index, unsigned int level,
RyoheiHagimoto 0:0e0631af0305 244 std::vector<lsh::BucketKey>& xor_masks)
RyoheiHagimoto 0:0e0631af0305 245 {
RyoheiHagimoto 0:0e0631af0305 246 xor_masks.push_back(key);
RyoheiHagimoto 0:0e0631af0305 247 if (level == 0) return;
RyoheiHagimoto 0:0e0631af0305 248 for (int index = lowest_index - 1; index >= 0; --index) {
RyoheiHagimoto 0:0e0631af0305 249 // Create a new key
RyoheiHagimoto 0:0e0631af0305 250 lsh::BucketKey new_key = key | (1 << index);
RyoheiHagimoto 0:0e0631af0305 251 fill_xor_mask(new_key, index, level - 1, xor_masks);
RyoheiHagimoto 0:0e0631af0305 252 }
RyoheiHagimoto 0:0e0631af0305 253 }
RyoheiHagimoto 0:0e0631af0305 254
RyoheiHagimoto 0:0e0631af0305 255 /** Performs the approximate nearest-neighbor search.
RyoheiHagimoto 0:0e0631af0305 256 * @param vec the feature to analyze
RyoheiHagimoto 0:0e0631af0305 257 * @param do_radius flag indicating if we check the radius too
RyoheiHagimoto 0:0e0631af0305 258 * @param radius the radius if it is a radius search
RyoheiHagimoto 0:0e0631af0305 259 * @param do_k flag indicating if we limit the number of nn
RyoheiHagimoto 0:0e0631af0305 260 * @param k_nn the number of nearest neighbors
RyoheiHagimoto 0:0e0631af0305 261 * @param checked_average used for debugging
RyoheiHagimoto 0:0e0631af0305 262 */
RyoheiHagimoto 0:0e0631af0305 263 void getNeighbors(const ElementType* vec, bool /*do_radius*/, float radius, bool do_k, unsigned int k_nn,
RyoheiHagimoto 0:0e0631af0305 264 float& /*checked_average*/)
RyoheiHagimoto 0:0e0631af0305 265 {
RyoheiHagimoto 0:0e0631af0305 266 static std::vector<ScoreIndexPair> score_index_heap;
RyoheiHagimoto 0:0e0631af0305 267
RyoheiHagimoto 0:0e0631af0305 268 if (do_k) {
RyoheiHagimoto 0:0e0631af0305 269 unsigned int worst_score = std::numeric_limits<unsigned int>::max();
RyoheiHagimoto 0:0e0631af0305 270 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table = tables_.begin();
RyoheiHagimoto 0:0e0631af0305 271 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table_end = tables_.end();
RyoheiHagimoto 0:0e0631af0305 272 for (; table != table_end; ++table) {
RyoheiHagimoto 0:0e0631af0305 273 size_t key = table->getKey(vec);
RyoheiHagimoto 0:0e0631af0305 274 std::vector<lsh::BucketKey>::const_iterator xor_mask = xor_masks_.begin();
RyoheiHagimoto 0:0e0631af0305 275 std::vector<lsh::BucketKey>::const_iterator xor_mask_end = xor_masks_.end();
RyoheiHagimoto 0:0e0631af0305 276 for (; xor_mask != xor_mask_end; ++xor_mask) {
RyoheiHagimoto 0:0e0631af0305 277 size_t sub_key = key ^ (*xor_mask);
RyoheiHagimoto 0:0e0631af0305 278 const lsh::Bucket* bucket = table->getBucketFromKey(sub_key);
RyoheiHagimoto 0:0e0631af0305 279 if (bucket == 0) continue;
RyoheiHagimoto 0:0e0631af0305 280
RyoheiHagimoto 0:0e0631af0305 281 // Go over each descriptor index
RyoheiHagimoto 0:0e0631af0305 282 std::vector<lsh::FeatureIndex>::const_iterator training_index = bucket->begin();
RyoheiHagimoto 0:0e0631af0305 283 std::vector<lsh::FeatureIndex>::const_iterator last_training_index = bucket->end();
RyoheiHagimoto 0:0e0631af0305 284 DistanceType hamming_distance;
RyoheiHagimoto 0:0e0631af0305 285
RyoheiHagimoto 0:0e0631af0305 286 // Process the rest of the candidates
RyoheiHagimoto 0:0e0631af0305 287 for (; training_index < last_training_index; ++training_index) {
RyoheiHagimoto 0:0e0631af0305 288 hamming_distance = distance_(vec, dataset_[*training_index], dataset_.cols);
RyoheiHagimoto 0:0e0631af0305 289
RyoheiHagimoto 0:0e0631af0305 290 if (hamming_distance < worst_score) {
RyoheiHagimoto 0:0e0631af0305 291 // Insert the new element
RyoheiHagimoto 0:0e0631af0305 292 score_index_heap.push_back(ScoreIndexPair(hamming_distance, training_index));
RyoheiHagimoto 0:0e0631af0305 293 std::push_heap(score_index_heap.begin(), score_index_heap.end());
RyoheiHagimoto 0:0e0631af0305 294
RyoheiHagimoto 0:0e0631af0305 295 if (score_index_heap.size() > (unsigned int)k_nn) {
RyoheiHagimoto 0:0e0631af0305 296 // Remove the highest distance value as we have too many elements
RyoheiHagimoto 0:0e0631af0305 297 std::pop_heap(score_index_heap.begin(), score_index_heap.end());
RyoheiHagimoto 0:0e0631af0305 298 score_index_heap.pop_back();
RyoheiHagimoto 0:0e0631af0305 299 // Keep track of the worst score
RyoheiHagimoto 0:0e0631af0305 300 worst_score = score_index_heap.front().first;
RyoheiHagimoto 0:0e0631af0305 301 }
RyoheiHagimoto 0:0e0631af0305 302 }
RyoheiHagimoto 0:0e0631af0305 303 }
RyoheiHagimoto 0:0e0631af0305 304 }
RyoheiHagimoto 0:0e0631af0305 305 }
RyoheiHagimoto 0:0e0631af0305 306 }
RyoheiHagimoto 0:0e0631af0305 307 else {
RyoheiHagimoto 0:0e0631af0305 308 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table = tables_.begin();
RyoheiHagimoto 0:0e0631af0305 309 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table_end = tables_.end();
RyoheiHagimoto 0:0e0631af0305 310 for (; table != table_end; ++table) {
RyoheiHagimoto 0:0e0631af0305 311 size_t key = table->getKey(vec);
RyoheiHagimoto 0:0e0631af0305 312 std::vector<lsh::BucketKey>::const_iterator xor_mask = xor_masks_.begin();
RyoheiHagimoto 0:0e0631af0305 313 std::vector<lsh::BucketKey>::const_iterator xor_mask_end = xor_masks_.end();
RyoheiHagimoto 0:0e0631af0305 314 for (; xor_mask != xor_mask_end; ++xor_mask) {
RyoheiHagimoto 0:0e0631af0305 315 size_t sub_key = key ^ (*xor_mask);
RyoheiHagimoto 0:0e0631af0305 316 const lsh::Bucket* bucket = table->getBucketFromKey(sub_key);
RyoheiHagimoto 0:0e0631af0305 317 if (bucket == 0) continue;
RyoheiHagimoto 0:0e0631af0305 318
RyoheiHagimoto 0:0e0631af0305 319 // Go over each descriptor index
RyoheiHagimoto 0:0e0631af0305 320 std::vector<lsh::FeatureIndex>::const_iterator training_index = bucket->begin();
RyoheiHagimoto 0:0e0631af0305 321 std::vector<lsh::FeatureIndex>::const_iterator last_training_index = bucket->end();
RyoheiHagimoto 0:0e0631af0305 322 DistanceType hamming_distance;
RyoheiHagimoto 0:0e0631af0305 323
RyoheiHagimoto 0:0e0631af0305 324 // Process the rest of the candidates
RyoheiHagimoto 0:0e0631af0305 325 for (; training_index < last_training_index; ++training_index) {
RyoheiHagimoto 0:0e0631af0305 326 // Compute the Hamming distance
RyoheiHagimoto 0:0e0631af0305 327 hamming_distance = distance_(vec, dataset_[*training_index], dataset_.cols);
RyoheiHagimoto 0:0e0631af0305 328 if (hamming_distance < radius) score_index_heap.push_back(ScoreIndexPair(hamming_distance, training_index));
RyoheiHagimoto 0:0e0631af0305 329 }
RyoheiHagimoto 0:0e0631af0305 330 }
RyoheiHagimoto 0:0e0631af0305 331 }
RyoheiHagimoto 0:0e0631af0305 332 }
RyoheiHagimoto 0:0e0631af0305 333 }
RyoheiHagimoto 0:0e0631af0305 334
RyoheiHagimoto 0:0e0631af0305 335 /** Performs the approximate nearest-neighbor search.
RyoheiHagimoto 0:0e0631af0305 336 * This is a slower version than the above as it uses the ResultSet
RyoheiHagimoto 0:0e0631af0305 337 * @param vec the feature to analyze
RyoheiHagimoto 0:0e0631af0305 338 */
RyoheiHagimoto 0:0e0631af0305 339 void getNeighbors(const ElementType* vec, ResultSet<DistanceType>& result)
RyoheiHagimoto 0:0e0631af0305 340 {
RyoheiHagimoto 0:0e0631af0305 341 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table = tables_.begin();
RyoheiHagimoto 0:0e0631af0305 342 typename std::vector<lsh::LshTable<ElementType> >::const_iterator table_end = tables_.end();
RyoheiHagimoto 0:0e0631af0305 343 for (; table != table_end; ++table) {
RyoheiHagimoto 0:0e0631af0305 344 size_t key = table->getKey(vec);
RyoheiHagimoto 0:0e0631af0305 345 std::vector<lsh::BucketKey>::const_iterator xor_mask = xor_masks_.begin();
RyoheiHagimoto 0:0e0631af0305 346 std::vector<lsh::BucketKey>::const_iterator xor_mask_end = xor_masks_.end();
RyoheiHagimoto 0:0e0631af0305 347 for (; xor_mask != xor_mask_end; ++xor_mask) {
RyoheiHagimoto 0:0e0631af0305 348 size_t sub_key = key ^ (*xor_mask);
RyoheiHagimoto 0:0e0631af0305 349 const lsh::Bucket* bucket = table->getBucketFromKey((lsh::BucketKey)sub_key);
RyoheiHagimoto 0:0e0631af0305 350 if (bucket == 0) continue;
RyoheiHagimoto 0:0e0631af0305 351
RyoheiHagimoto 0:0e0631af0305 352 // Go over each descriptor index
RyoheiHagimoto 0:0e0631af0305 353 std::vector<lsh::FeatureIndex>::const_iterator training_index = bucket->begin();
RyoheiHagimoto 0:0e0631af0305 354 std::vector<lsh::FeatureIndex>::const_iterator last_training_index = bucket->end();
RyoheiHagimoto 0:0e0631af0305 355 DistanceType hamming_distance;
RyoheiHagimoto 0:0e0631af0305 356
RyoheiHagimoto 0:0e0631af0305 357 // Process the rest of the candidates
RyoheiHagimoto 0:0e0631af0305 358 for (; training_index < last_training_index; ++training_index) {
RyoheiHagimoto 0:0e0631af0305 359 // Compute the Hamming distance
RyoheiHagimoto 0:0e0631af0305 360 hamming_distance = distance_(vec, dataset_[*training_index], (int)dataset_.cols);
RyoheiHagimoto 0:0e0631af0305 361 result.addPoint(hamming_distance, *training_index);
RyoheiHagimoto 0:0e0631af0305 362 }
RyoheiHagimoto 0:0e0631af0305 363 }
RyoheiHagimoto 0:0e0631af0305 364 }
RyoheiHagimoto 0:0e0631af0305 365 }
RyoheiHagimoto 0:0e0631af0305 366
RyoheiHagimoto 0:0e0631af0305 367 /** The different hash tables */
RyoheiHagimoto 0:0e0631af0305 368 std::vector<lsh::LshTable<ElementType> > tables_;
RyoheiHagimoto 0:0e0631af0305 369
RyoheiHagimoto 0:0e0631af0305 370 /** The data the LSH tables where built from */
RyoheiHagimoto 0:0e0631af0305 371 Matrix<ElementType> dataset_;
RyoheiHagimoto 0:0e0631af0305 372
RyoheiHagimoto 0:0e0631af0305 373 /** The size of the features (as ElementType[]) */
RyoheiHagimoto 0:0e0631af0305 374 unsigned int feature_size_;
RyoheiHagimoto 0:0e0631af0305 375
RyoheiHagimoto 0:0e0631af0305 376 IndexParams index_params_;
RyoheiHagimoto 0:0e0631af0305 377
RyoheiHagimoto 0:0e0631af0305 378 /** table number */
RyoheiHagimoto 0:0e0631af0305 379 unsigned int table_number_;
RyoheiHagimoto 0:0e0631af0305 380 /** key size */
RyoheiHagimoto 0:0e0631af0305 381 unsigned int key_size_;
RyoheiHagimoto 0:0e0631af0305 382 /** How far should we look for neighbors in multi-probe LSH */
RyoheiHagimoto 0:0e0631af0305 383 unsigned int multi_probe_level_;
RyoheiHagimoto 0:0e0631af0305 384
RyoheiHagimoto 0:0e0631af0305 385 /** The XOR masks to apply to a key to get the neighboring buckets */
RyoheiHagimoto 0:0e0631af0305 386 std::vector<lsh::BucketKey> xor_masks_;
RyoheiHagimoto 0:0e0631af0305 387
RyoheiHagimoto 0:0e0631af0305 388 Distance distance_;
RyoheiHagimoto 0:0e0631af0305 389 };
RyoheiHagimoto 0:0e0631af0305 390 }
RyoheiHagimoto 0:0e0631af0305 391
RyoheiHagimoto 0:0e0631af0305 392 #endif //OPENCV_FLANN_LSH_INDEX_H_