| 1 | /* ---------------------------------------------------------------------------- |
| 2 | |
| 3 | * GTSAM Copyright 2010, Georgia Tech Research Corporation, |
| 4 | * Atlanta, Georgia 30332-0415 |
| 5 | * All Rights Reserved |
| 6 | * Authors: Frank Dellaert, et al. (see THANKS for the full author list) |
| 7 | |
| 8 | * See LICENSE for the license information |
| 9 | |
| 10 | * -------------------------------------------------------------------------- */ |
| 11 | |
| 12 | /** |
| 13 | * @file SFMExample_bal_COLAMD_METIS.cpp |
| 14 | * @brief This file is to compare the ordering performance for COLAMD vs METIS. |
| 15 | * Example problem is to solve a structure-from-motion problem from a "Bundle Adjustment in the Large" file. |
| 16 | * @author Frank Dellaert, Zhaoyang Lv |
| 17 | */ |
| 18 | |
| 19 | // For an explanation of headers, see SFMExample.cpp |
| 20 | #include <gtsam/slam/GeneralSFMFactor.h> |
| 21 | #include <gtsam/sfm/SfmData.h> // for loading BAL datasets ! |
| 22 | #include <gtsam/slam/dataset.h> |
| 23 | #include <gtsam/nonlinear/NonlinearFactorGraph.h> |
| 24 | #include <gtsam/nonlinear/LevenbergMarquardtOptimizer.h> |
| 25 | #include <gtsam/inference/Symbol.h> |
| 26 | #include <gtsam/inference/Ordering.h> |
| 27 | #include <gtsam/base/timing.h> |
| 28 | |
| 29 | #include <vector> |
| 30 | |
| 31 | using namespace std; |
| 32 | using namespace gtsam; |
| 33 | using symbol_shorthand::C; |
| 34 | using symbol_shorthand::P; |
| 35 | |
| 36 | // We will be using a projection factor that ties a SFM_Camera to a 3D point. |
| 37 | // An SFM_Camera is defined in datase.h as a camera with unknown Cal3Bundler |
| 38 | // calibration and has a total of 9 free parameters |
| 39 | typedef GeneralSFMFactor<SfmCamera, Point3> MyFactor; |
| 40 | |
| 41 | |
| 42 | int main(int argc, char* argv[]) { |
| 43 | // Find default file, but if an argument is given, try loading a file |
| 44 | string filename = findExampleDataFile(name: "dubrovnik-3-7-pre" ); |
| 45 | if (argc > 1) filename = string(argv[1]); |
| 46 | |
| 47 | // Load the SfM data from file |
| 48 | SfmData mydata = SfmData::FromBalFile(filename); |
| 49 | cout << "read " << mydata.numberTracks() << " tracks on " << mydata.numberCameras() << " cameras" << endl; |
| 50 | |
| 51 | // Create a factor graph |
| 52 | NonlinearFactorGraph graph; |
| 53 | |
| 54 | // We share *one* noiseModel between all projection factors |
| 55 | auto noise = noiseModel::Isotropic::Sigma(dim: 2, sigma: 1.0); // one pixel in u and v |
| 56 | |
| 57 | // Add measurements to the factor graph |
| 58 | size_t j = 0; |
| 59 | for (const SfmTrack& track : mydata.tracks) { |
| 60 | for (const auto& [i, uv] : track.measurements) { |
| 61 | graph.emplace_shared<MyFactor>( |
| 62 | args: uv, args&: noise, args: C(j: i), args: P(j)); // note use of shorthand symbols C and P |
| 63 | } |
| 64 | j += 1; |
| 65 | } |
| 66 | |
| 67 | // Add a prior on pose x1. This indirectly specifies where the origin is. |
| 68 | // and a prior on the position of the first landmark to fix the scale |
| 69 | graph.addPrior(key: C(j: 0), prior: mydata.cameras[0], model: noiseModel::Isotropic::Sigma(dim: 9, sigma: 0.1)); |
| 70 | graph.addPrior(key: P(j: 0), prior: mydata.tracks[0].p, |
| 71 | model: noiseModel::Isotropic::Sigma(dim: 3, sigma: 0.1)); |
| 72 | |
| 73 | // Create initial estimate |
| 74 | Values initial; |
| 75 | size_t i = 0; |
| 76 | j = 0; |
| 77 | for (const SfmCamera& camera : mydata.cameras) initial.insert(j: C(j: i++), val: camera); |
| 78 | for (const SfmTrack& track : mydata.tracks) initial.insert(j: P(j: j++), val: track.p); |
| 79 | |
| 80 | /** --------------- COMPARISON -----------------------**/ |
| 81 | /** ----------------------------------------------------**/ |
| 82 | |
| 83 | LevenbergMarquardtParams params_using_COLAMD, params_using_METIS; |
| 84 | try { |
| 85 | params_using_METIS.setVerbosity("ERROR" ); |
| 86 | gttic_(METIS_ORDERING); |
| 87 | params_using_METIS.ordering = Ordering::Create(orderingType: Ordering::METIS, graph); |
| 88 | gttoc_(METIS_ORDERING); |
| 89 | |
| 90 | params_using_COLAMD.setVerbosity("ERROR" ); |
| 91 | gttic_(COLAMD_ORDERING); |
| 92 | params_using_COLAMD.ordering = Ordering::Create(orderingType: Ordering::COLAMD, graph); |
| 93 | gttoc_(COLAMD_ORDERING); |
| 94 | } catch (exception& e) { |
| 95 | cout << e.what(); |
| 96 | } |
| 97 | |
| 98 | // expect they have different ordering results |
| 99 | if (params_using_COLAMD.ordering == params_using_METIS.ordering) { |
| 100 | cout << "COLAMD and METIS produce the same ordering. " |
| 101 | << "Problem here!!!" << endl; |
| 102 | } |
| 103 | |
| 104 | /* Optimize the graph with METIS and COLAMD and time the results */ |
| 105 | |
| 106 | Values result_METIS, result_COLAMD; |
| 107 | try { |
| 108 | gttic_(OPTIMIZE_WITH_METIS); |
| 109 | LevenbergMarquardtOptimizer lm_METIS(graph, initial, params_using_METIS); |
| 110 | result_METIS = lm_METIS.optimize(); |
| 111 | gttoc_(OPTIMIZE_WITH_METIS); |
| 112 | |
| 113 | gttic_(OPTIMIZE_WITH_COLAMD); |
| 114 | LevenbergMarquardtOptimizer lm_COLAMD(graph, initial, params_using_COLAMD); |
| 115 | result_COLAMD = lm_COLAMD.optimize(); |
| 116 | gttoc_(OPTIMIZE_WITH_COLAMD); |
| 117 | } catch (exception& e) { |
| 118 | cout << e.what(); |
| 119 | } |
| 120 | |
| 121 | { // printing the result |
| 122 | |
| 123 | cout << "COLAMD final error: " << graph.error(values: result_COLAMD) << endl; |
| 124 | cout << "METIS final error: " << graph.error(values: result_METIS) << endl; |
| 125 | |
| 126 | cout << endl << endl; |
| 127 | |
| 128 | cout << "Time comparison by solving " << filename << " results:" << endl; |
| 129 | |
| 130 | cout << mydata.numberTracks() << " point tracks and " << mydata.numberCameras() |
| 131 | << " cameras" << endl; |
| 132 | |
| 133 | tictoc_print_(); |
| 134 | } |
| 135 | |
| 136 | return 0; |
| 137 | } |
| 138 | |
| 139 | |