PoSDK Data Converters

This document details various data converters provided by PoSDK for data format conversion between PoSDK and other mainstream computer vision libraries/tools.

---

Converter Module Structure

PoSDK converters adopt a modular design, with each converter responsible for data interaction with specific external libraries or tools.

File Organization

common/converter/
├── converter_base.hpp         # Converter base class
├── converter_opencv.hpp/cpp    # OpenCV data converter
├── converter_openmvg_file.hpp/cpp # OpenMVG file converter
├── converter_opengv.hpp/cpp    # OpenGV data converter
├── converter_colmap_file.hpp/cpp  # COLMAP file converter
└── CMakeLists.txt             # Build configuration

Module Responsibilities

Module	Main Functions	Documentation Link
OpenCVConverter	Feature point conversion, Descriptor conversion, Match conversion, Camera model conversion	OpenCV Converter
OpenMVGFileConverter	SfM data loading, Feature file conversion, Match file conversion	OpenMVG Converter
OpenGVConverter	Bearing vector conversion, Pose conversion, Camera parameter conversion	OpenGV Converter
ColmapConverter	SfM data conversion, Match data conversion, Data export	COLMAP Converter

---

Converter Feature Comparison

Conversion Feature	OpenCV	OpenMVG	OpenGV	COLMAP
Feature Data
Feature points	✓ Bidirectional conversion	✓ File reading	-	-
Descriptors	✓ Bidirectional conversion	-	-	-
Matches	✓ Bidirectional conversion	✓ File reading	✓ Match to Bearing	✓ File reading
Geometric Data
Bearing vectors	-	-	✓ Pixel/match conversion	-
Camera models	✓ Calibration conversion	✓ SfM file reading	✓ Intrinsic extraction	✓ File reading/export
Relative poses	-	-	✓ Algorithm result conversion	-
Global poses	-	✓ SfM file reading	-	✓ File reading
3D Data
3D point clouds	-	✓ SfM file reading	-	✓ File reading/export
Other Data
Image paths	-	✓ SfM file reading	-	-
PLY export	-	-	-	✓ Point cloud/scene export

Legend:

• ✓ : Feature supported

• - : Not supported

---

Detailed Documentation

For converter usage instructions, API reference, and example code, please refer to the following sub-documents:

• OpenCV Data Converter
  • Conversion Architecture
  • Feature Point Conversion
  • Descriptor Conversion
  • Image Feature Information Conversion
  • Batch Feature Conversion
  • Match Conversion
  • Camera Model Conversion
  • Error Handling and Debugging
  • Complete Workflow Example
• OpenMVG File Converter
  • Conversion Function Overview
  • File Loading Functions
  • Data Conversion Functions
  • Complete Workflow Examples
  • OpenMVG File Format Description
  • Error Handling
• OpenGV Data Converter
  • Conversion Function Overview
  • Bearing Vector Conversion
  • Pose Conversion
  • Camera Parameter Conversion
  • Complete Workflow Examples
  • OpenGV Algorithm Support
  • Error Handling
• COLMAP File Converter
  • Conversion Function Overview
  • COLMAP → PoSDK Conversion
  • PoSDK → COLMAP Export
  • COLMAP File Format Support
  • Coordinate System Conventions
  • Complete Workflow Examples
  • COLMAP Binary Format Description
  • Error Handling

---

Quick Start Examples

OpenCV Feature Extraction and Conversion

#include <opencv2/features2d.hpp>
#include "converter_opencv.hpp"

using namespace PoSDK;
using namespace PoSDK::Converter;

// OpenCV feature extraction
cv::Ptr<cv::SIFT> detector = cv::SIFT::create();
std::vector<cv::KeyPoint> keypoints;
cv::Mat descriptors;
detector->detectAndCompute(image, cv::noArray(), keypoints, descriptors);

// Convert to PoSDK format
ImageFeatureInfo image_features;
Descs descs;
OpenCVConverter::CVFeatures2ImageFeatureInfoWithDesc(
    keypoints, descriptors, image_features, descs, 
    image_path, "SIFT");

OpenMVG Data Loading

#include "converter_openmvg_file.hpp"

using namespace PoSDK::Converter;

// Load OpenMVG SfM results
Interface::DataPtr global_poses_data;
OpenMVGFileConverter::ToDataGlobalPoses(
    "sfm_data.json", global_poses_data);

OpenGV Pose Estimation

#include "converter_opengv.hpp"

using namespace PoSDK::Converter;

// Convert matches to Bearing vectors
opengv::bearingVectors_t bearings1, bearings2;
OpenGVConverter::MatchesToBearingVectors(
    matches, features_info, camera_models, view_pair,
    bearings1, bearings2);

COLMAP Data Loading

#include "converter_colmap_file.hpp"

using namespace PoSDK::Converter::COLMAP;

// 1. Create file name to ID mapping (from OpenMVG SfM file)
std::map<std::string, int> file_name_to_id;
SfMFileToIdMap("sfm_data.json", file_name_to_id);

// 2. Load global poses (from COLMAP images.txt file)
types::GlobalPoses global_poses;
Interface::DataPtr global_poses_data = 
    std::make_shared<Interface::DataMap<types::GlobalPoses>>(global_poses);
ToDataGlobalPoses("images.txt", global_poses_data, file_name_to_id);

// 3. Load match data (from COLMAP matches folder)
types::Matches matches;
Interface::DataPtr matches_data = 
    std::make_shared<Interface::DataMap<types::Matches>>(matches);
ToDataMatches("matches/", matches_data, file_name_to_id);

Export to COLMAP Data Format

#include "converter_colmap_file.hpp"

using namespace PoSDK::Converter::COLMAP;

// Export PoSDK results to COLMAP format
OutputPoSDK2Colmap(output_path, global_poses, camera_models,
                   features, tracks, pts3d);

---

Related Documentation:

• Core Data Types - Learn PoSDK data type definitions

• Plugin Development - Learn how to develop custom plugins

• Advanced Features - Explore advanced development features