Skip to content

Replace colmap flann with nanoflann for knnSearch #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
yuanxion wants to merge 3 commits into
base: dev-colmap
Choose a base branch
from
Draft

Replace colmap flann with nanoflann for knnSearch #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3ee83a4
Using nanoflann to do the knnSearch.
yuanxion Mar 16, 2023
b7b5cab
Compare flann and nanoflann search distance results.
yuanxion Mar 16, 2023
35200c8
Compare flann and nanoflann running time.
yuanxion Mar 16, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion src/feature/matching.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -658,7 +658,8 @@ SiftFeatureMatcher::SiftFeatureMatcher(const SiftMatchingOptions& options,
: options_(options), database_(database), cache_(cache), is_setup_(false) {
CHECK(options_.Check());

const int num_threads = GetEffectiveNumThreads(options_.num_threads);
//const int num_threads = GetEffectiveNumThreads(options_.num_threads);
const int num_threads = 1;
CHECK_GT(num_threads, 0);

std::vector<int> gpu_indices = CSVToVector<int>(options_.gpu_index);
Expand Down Expand Up @@ -907,6 +908,7 @@ void ExhaustiveFeatureMatcher::Run() {
const std::vector<image_t> image_ids = cache_.GetImageIds();

const size_t block_size = static_cast<size_t>(options_.block_size);
//const size_t block_size(53);
const size_t num_blocks = static_cast<size_t>(
std::ceil(static_cast<double>(image_ids.size()) / block_size));
const size_t num_pairs_per_block = block_size * (block_size - 1) / 2;
Expand Down
88 changes: 88 additions & 0 deletions src/feature/sift.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,8 @@
#include "util/misc.h"
#include "util/opengl_utils.h"

#include "util/nanoflann.hpp"

namespace colmap {
namespace {

Expand Down Expand Up @@ -203,6 +205,15 @@ Eigen::MatrixXi ComputeSiftDistanceMatrix(
return dists;
}

template<typename T>
void myprint(T* vec) {
const int N =16;
for(int i=0; i<N; ++i) {
std::cout << vec[i] << " ";
}
std::cout << std::endl;
}

void FindNearestNeighborsFLANN(
const FeatureDescriptors& query, const FeatureDescriptors& database,
Eigen::Matrix<int, Eigen::Dynamic, Eigen::Dynamic, Eigen::RowMajor>*
Expand All @@ -228,6 +239,10 @@ void FindNearestNeighborsFLANN(

flann::Matrix<int> indices_matrix(indices->data(), query.rows(),
num_nearest_neighbors);
#if 1
Timer timer;
timer.Start();

std::vector<float> distances_vector(query.rows() * num_nearest_neighbors);
flann::Matrix<float> distances_matrix(distances_vector.data(), query.rows(),
num_nearest_neighbors);
Expand All @@ -236,17 +251,86 @@ void FindNearestNeighborsFLANN(
index.buildIndex();
index.knnSearch(query_matrix, indices_matrix, distances_matrix,
num_nearest_neighbors, flann::SearchParams(128));
std::cout << "flann ------------------------------------------------- " << StringPrintf(" in %.3fs", timer.ElapsedSeconds()) << std::endl;

for (Eigen::Index query_index = 0; query_index < indices->rows();
++query_index) {
if(query_index < 4) {
auto q = query.row(query_index).cast<int>();
std::cout << "q ";
for(int i=0; i<48; ++i) std::cout << q[i] << " ";
std::cout << "..." << std::endl;
}
for (Eigen::Index k = 0; k < indices->cols(); ++k) {
const Eigen::Index database_index = indices->coeff(query_index, k);
distances->coeffRef(query_index, k) =
query.row(query_index)
.cast<int>()
.dot(database.row(database_index).cast<int>());
if(query_index < 4) {
std::cout << "d ";
auto d = database.row(database_index).cast<int>();
for(int i=0; i<48; ++i) std::cout << d[i] << " ";
std::cout << "... \t[index:" << database_index << " distance:" << distances->coeffRef(query_index, k) << "]" << std::endl;
}
}
}
//#else
timer.Restart();
using my_kd_tree_t = nanoflann::KDTreeEigenMatrixAdaptor<FeatureDescriptors>;
std::vector<uint8_t> distances_vector_u8(query.rows() * num_nearest_neighbors);
flann::Matrix<uint8_t> distances_matrix_u8(distances_vector_u8.data(), query.rows(),
num_nearest_neighbors);
const size_t dim(128);
const int max_leaf = 10;
const int query_rows = query.rows();
my_kd_tree_t mat_index(dim, std::cref(database), max_leaf);

for (int i = 0; i < query_rows; ++i) {
uint8_t* query_vec = query_matrix[i];
uint8_t* distances_ptr = distances_matrix_u8[i];
int* indices_ptr = indices_matrix[i];

std::vector<long int> ret_indexes(kNumNearestNeighbors);
std::vector<uint8_t> out_dist_sqr(kNumNearestNeighbors);
size_t res = mat_index.index_->knnSearch(query_vec, kNumNearestNeighbors, &ret_indexes[0], &out_dist_sqr[0]);

size_t finds = std::min(res, kNumNearestNeighbors);
for (int j = 0; j < finds; ++j) {
*indices_ptr++ = ret_indexes[j];
//*distances_ptr++ = out_dist_sqr[j];
//if(i < 8)
// std::cout << __FILE__ << ":" << __LINE__ << " query " << i << "/" << query_rows << " finds " << j << "/" << finds << " index " << ret_indexes[j] << " distance " << (int)out_dist_sqr[j] << std::endl;
}
}
std::cout << "nanoflann ------------------------------------------------- " << StringPrintf(" in %.3fs", timer.ElapsedSeconds()) << std::endl;

for (Eigen::Index query_index = 0; query_index < indices->rows();
++query_index) {
if(query_index < 4) {
auto q = query.row(query_index).cast<int>();
std::cout << "q ";
for(int i=0; i<48; ++i) std::cout << q[i] << " ";
std::cout << "..." << std::endl;
}
for (Eigen::Index k = 0; k < indices->cols(); ++k) {
const Eigen::Index database_index = indices->coeff(query_index, k);
//if(query_index < 8)
// std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " indices " << query_index << "," << k << " database_index " << database_index << std::endl;
distances->coeffRef(query_index, k) =
query.row(query_index)
.cast<int>()
.dot(database.row(database_index).cast<int>());
if(query_index < 4) {
std::cout << "d ";
auto d = database.row(database_index).cast<int>();
for(int i=0; i<48; ++i) std::cout << d[i] << " ";
std::cout << "... \t[index:" << database_index << " distance:" << distances->coeffRef(query_index, k) << "]" << std::endl;
}
}
}
#endif
//std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " " << __func__ << " query " << query.rows() << " finished" << std::endl;
}

size_t FindBestMatchesOneWayFLANN(
Expand Down Expand Up @@ -319,6 +403,7 @@ void FindBestMatchesFLANN(
const float max_ratio, const float max_distance, const bool cross_check,
FeatureMatches* matches) {
matches->clear();
//std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " FindBestMatchesFLANN" << std::endl;

std::vector<int> matches12;
const size_t num_matches12 = FindBestMatchesOneWayFLANN(
Expand Down Expand Up @@ -992,17 +1077,20 @@ void MatchSiftFeaturesCPUFLANN(const SiftMatchingOptions& match_options,
Eigen::Matrix<int, Eigen::Dynamic, Eigen::Dynamic, Eigen::RowMajor>
distances_2to1;

//std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " start FindNearestNeighborsFLANN" << std::endl;
FindNearestNeighborsFLANN(descriptors1, descriptors2, &indices_1to2,
&distances_1to2);
if (match_options.cross_check) {
FindNearestNeighborsFLANN(descriptors2, descriptors1, &indices_2to1,
&distances_2to1);
}

//std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " start FindBestMatchesFLANN" << std::endl;
FindBestMatchesFLANN(indices_1to2, distances_1to2, indices_2to1,
distances_2to1, match_options.max_ratio,
match_options.max_distance, match_options.cross_check,
matches);
//std::cout << "[xy] " << __FILE__ << ":" << __LINE__ << " start FindBestMatchesFLANN finished" << std::endl;
}

void MatchSiftFeaturesCPU(const SiftMatchingOptions& match_options,
Expand Down
Loading