update readme.txt

CMakeLists.txt

@@ -171,20 +171,20 @@ endif()
 add_executable(run_matching src/run_matching.cc)
 target_link_libraries(run_matching xrsfm)
 
-# add_executable(run_reconstruction src/run_reconstruction.cc)
-# target_link_libraries(run_reconstruction xrsfm)
+add_executable(run_reconstruction src/run_reconstruction.cc)
+target_link_libraries(run_reconstruction xrsfm)
 
-# add_executable(run_triangulation src/run_triangulation.cc)
-# target_link_libraries(run_triangulation xrsfm)
+add_executable(run_triangulation src/run_triangulation.cc)
+target_link_libraries(run_triangulation xrsfm)
 
-# add_executable(unpack_collect_data src/unpack_collect_data.cc)
-# target_link_libraries(unpack_collect_data xrsfm)
+add_executable(unpack_collect_data src/unpack_collect_data.cc)
+target_link_libraries(unpack_collect_data xrsfm)
 
-# add_executable(estimate_scale src/estimate_scale.cc)
-# target_link_libraries(estimate_scale xrsfm apriltag)
+add_executable(estimate_scale src/estimate_scale.cc)
+target_link_libraries(estimate_scale xrsfm apriltag)
 
-# add_executable(rec_1dsfm src/rec_1dsfm.cc)
-# target_link_libraries(rec_1dsfm xrsfm)
+add_executable(rec_1dsfm src/rec_1dsfm.cc)
+target_link_libraries(rec_1dsfm xrsfm)
 
 add_executable(rec_kitti src/rec_kitti.cc)
 target_link_libraries(rec_kitti xrsfm)

Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:10.2-devel-ubuntu18.04
+FROM nvidia/cuda:12.1.0-devel-ubuntu20.04
 
 ENV DEBIAN_FRONTEND=noninteractive
 

docs/en/tutorial.md

@@ -51,7 +51,7 @@ Run reconsturction stage with the following command line
 "init_id1" and "init_id2" indicates two frames for the initialization of SfM.
 In general, these two frames can be set to 0 and 5.
 
-The format of output binary files is is consistent with COLMAP, you can use colmap gui to view the reconstruction result.
+You can view the reconstruction results using ```./bin/xrsfm_gui```. The file format is the same as the well-known open-source project COLMAP. You can also use COLMAP's graphical user interface to view the reconstruction results of this project.
 
 ### 3. Estimate scale with apriltag (Optional)
 Input: images, reconstruction results

docs/zh/tutorial.md

@@ -49,7 +49,7 @@ XRSfM支持顺序匹配("sequential")、基于检索的匹配("retrieval")和基
 "init_id1" and "init_id2" 是可选的输入参数，指定重建阶段的初始帧，当不输入该参数时，程序将会自行选择初始帧。
 
 输出的重建结果由三部分组成“cameras.bin”,“images.bin”,"points.bin"。
-文件格式与知名的开源工作COLMAP相同，用户可以使用COLMAP的图像用户界面查看本项目的重建结果。
+使用```./bin/xrsfm_gui```可以查看重建结果，文件格式与知名的开源工作COLMAP相同，也可以使用COLMAP的图像用户界面查看本项目的重建结果。
 
 ### 3. 尺度估计（可选）
 输入: 图像文件夹，重建结果

src/base/camera.hpp

@@ -36,11 +36,8 @@ class Camera {
     }
     Camera(int _id, double fxy, double cx, double cy) {
         id_ = _id;
-        // model_id_ = 0;
-        // params_ = {fxy, cx, cy};
         model_id_ = 2;
         params_ = {fxy, cx, cy, 0};
-        std::cout << params_.size() << std::endl;
         is_valid = true;
     }
 
@@ -88,8 +85,8 @@ inline void ImageToNormalized(const Camera &c, const Eigen::Vector2d &p2d,
     CAMERA_MODEL_CASES
 #undef CAMERA_MODEL_CASE
 
-    std::cout << "ERROR: (ImageToNormalized) unsupported model type: "
-              << c.model_id_ << "\n";
+    printf("ERROR: (ImageToNormalized) unsupported model type: %d\n",
+           c.model_id_);
     exit(-1);
 }
 
@@ -104,8 +101,8 @@ inline void NormalizedToImage(const Camera &c, const Eigen::Vector2d &p2d_n,
     CAMERA_MODEL_CASES
 #undef CAMERA_MODEL_CASE
 
-    std::cout << "ERROR: (NormalizedToImage) unsupported model type: "
-              << c.model_id_ << "\n";
+    printf("ERROR: (NormalizedToImage) unsupported model type: %d",
+           int(c.model_id_));
     exit(-1);
 }
 

src/base/camera_model.hpp

@@ -26,7 +26,6 @@ inline void IterativeUndistortion(const double *params, double *xy) {
     Eigen::Vector2d xy_plus;
 
     for (size_t i = 0; i < kNumIterations; ++i) {
-        // std::cout << i << " " << kNumIterations << std::endl;
         const double step0 = std::max(std::numeric_limits<double>::epsilon(),
                                       std::abs(kRelStepSize * x(0)));
         const double step1 = std::max(std::numeric_limits<double>::epsilon(),

src/colmap/ransac/combination_sampler.cc

@@ -34,7 +34,7 @@
 #include <numeric>
 
 #include "random.h"
-#include "util/math.h"
+#include "math.h"
 
 namespace colmap {
 

src/colmap/ui/model_viewer_widget.cc

@@ -31,7 +31,6 @@
 
 #include <Eigen/Eigen>
 #include "model_viewer_widget.h"
-#include "util/math.h"
 
 #define SELECTION_BUFFER_IMAGE_IDX 0
 #define SELECTION_BUFFER_POINT_IDX 1
@@ -52,6 +51,10 @@ const Eigen::Vector4f kZAxisColor(0.0f, 0.0f, 0.9f, 0.5f);
 namespace xrsfm {
 namespace {
 
+double RadToDeg(const double rad) {
+    return rad * 57.29577951308232286464772187173366546630859375;
+}
+
 // Generate unique index from RGB color in the range [0, 256^3].
 inline size_t RGBToIndex(const uint8_t r, const uint8_t g, const uint8_t b) {
     return static_cast<size_t>(r) + static_cast<size_t>(g) * 256 +
@@ -369,8 +372,7 @@ void ModelViewerWidget::RotateView(const float x, const float y,
         // First shift to rotation center, then rotate and shift back.
         model_view_matrix_.translate(rot_center(0), rot_center(1),
                                      rot_center(2));
-        model_view_matrix_.rotate(colmap::RadToDeg(angle), axis(0), axis(1),
-                                  axis(2));
+        model_view_matrix_.rotate(RadToDeg(angle), axis(0), axis(1), axis(2));
         model_view_matrix_.translate(-rot_center(0), -rot_center(1),
                                      -rot_center(2));
         update();

src/estimate_scale.cc

@@ -3,7 +3,7 @@
 
 #include <fstream>
 
-#include "base/camera.h"
+#include "base/camera.hpp"
 #include "base/map.h"
 #include "geometry/track_processor.h"
 #include "optimization/ba_solver.h"

src/feature/feature_extraction.cc

@@ -40,17 +40,15 @@ void FeatureExtract(const std::string &image_dir_path,
         const cv::Mat image = cv::imread(image_dir_path + frame.name,
                                          cv::IMREAD_IGNORE_ORIENTATION);
         if (image.rows == 0) {
-            std::cout << "Can't read " << image_dir_path + frame.name
-                      << std::endl;
+            std::cout << "ERROR: fail to read image: "
+                      << image_dir_path + frame.name << std::endl;
             exit(0);
         }
 
         image_size.push_back(ImageSize(image.cols, image.rows));
 
 #ifndef USE_ORB
-        // std::cout << image_dir_path + frame.name << std::endl;
         sift.ExtractUINT8(image, frame.keypoints_, frame.uint_descs_);
-        // std::cout << i << " " << frame.uint_descs_.rows() << std::endl;
 #else
         orb(image, cv::Mat(), frame.keypoints_, frame.orb_descs_);
 #endif

src/geometry/error_detector.cc

@@ -92,9 +92,6 @@ bool ErrorDetector::IsGoodRelativePose(const Map &map, const FramePair &fp,
            num_matches);
 
     if (ratio < ratio_th) {
-        // std::cout << frame1.tcw.q.coeffs().transpose() << " " <<
-        // frame1.tcw.t.transpose() << std::endl;nn
-
         return false;
     }
     return true;

src/geometry/track_processor.cc

@@ -175,8 +175,8 @@ int Point3dProcessor::TriangulateFramePoint(Map &map, const int frame_id,
             if (!cor_frame.registered)
                 continue;
             if (cor_frame.track_ids_.size() <= t_p2d_id) {
-                std::cout << "bad" << t_frame_id << " " << t_p2d_id
-                          << std::endl;
+                printf("Error: frame %d index out of bounds %d %d\n",
+                       t_frame_id, t_p2d_id, int(cor_frame.track_ids_.size()));
                 exit(0);
             }
             observations.emplace_back(t_frame_id, t_p2d_id);

src/optimization/ba_solver.cc

@@ -5,12 +5,29 @@
 #include "ba_solver.h"
 
 #include "cost_factor_ceres.h"
-#include "util/math.h"
 #include "utility/timer.h"
 #include "geometry/triangulate_light.h"
 
 namespace xrsfm {
 
+template <typename T>
+T Percentile(const std::vector<T> &elems, const double p) {
+    CHECK(!elems.empty());
+    CHECK_GE(p, 0);
+    CHECK_LE(p, 100);
+
+    const int idx = static_cast<int>(std::round(p / 100 * (elems.size() - 1)));
+    const size_t percentile_idx =
+        std::max(0, std::min(static_cast<int>(elems.size() - 1), idx));
+
+    std::vector<T> ordered_elems = elems;
+    std::nth_element(ordered_elems.begin(),
+                     ordered_elems.begin() + percentile_idx,
+                     ordered_elems.end());
+
+    return ordered_elems.at(percentile_idx);
+}
+
 void PrintSolverSummary(const ceres::Solver::Summary &summary) {
     std::cout << std::right << std::setw(16) << "Residuals : ";
     std::cout << std::left << summary.num_residuals_reduced << std::endl;
@@ -466,7 +483,7 @@ std::vector<int> FindLocalBundle(const int frame_id, Map &map,
 
                 // Calculate the triangulation angle at a certain percentile.
                 const double kTriangulationAnglePercentile = 75;
-                tri_angle = colmap::Percentile(
+                tri_angle = Percentile(
                     CalculateTriangulationAnglesLight(
                         proj_center, proj_center_overlap, shared_points3D),
                     kTriangulationAnglePercentile);

src/rec_1dsfm.cc

@@ -41,7 +41,6 @@ void PreProcess(const std::string bin_path, Map &map) {
     }
 
     for (int i = 0; i < cameras.size(); ++i) {
-        std::cout << i << " " << cameras.size() << std::endl;
         auto &camera = cameras[i];
         // camera.log();
         // if distortion parameters of the camera are not estimated, the camera

src/tag/tag_extract.hpp

@@ -20,7 +20,7 @@ extern "C" {
 }
 
 #include "base/map.h"
-#include "base/camera.h"
+#include "base/camera.hpp"
 #include "geometry/track_processor.h"
 #include "optimization/ba_solver.h"
 #include "optimization/cost_factor_ceres.h"
@@ -257,7 +257,7 @@ void tag_refine(std::string image_dir, std::string map_dir,
 
     ceres::Solve(solver_options, &problem, &summary);
     std::cout << summary.BriefReport() << "\n";
-    std::cout << scale << std::endl;
+    std::cout << "scale :" << scale << std::endl;
 
     // resize map
     for (auto &[id, frame] : map.frame_map_) {