kenkenjlab/vitamin-e

Unofficial C++ implementation of SLAM method, VITAMIN-E

Unofficial implemnetation of VITAMIN-E (WIP)

• VITAMIN-E : VIsual Tracking And MappINg with Extremely Dense Feature Points (CVPR2019)
  • https://staff.aist.go.jp/shuji.oishi/assets/projects/VITAMIN-E/index.html
• NOTE
  • This project is ongoing, not done yet.
  • Not optimized yet and runs slowly.

---

Prerequisites

Please install the following tools and libraries:

• CMake 3.10+
• C++ Compiler (C++11 supports) such as:
  • Visual C++ 2015+ (if Windows)
  • GCC 7+ (if Linux)
• OpenCV 4+ with VTK
  • core
  • highgui
  • calib3d
  • imgproc
  • features2d
  • viz (VTK required)
  • video

This program is confirmed running successfully in the following environments:

• Windows
  • Windows 10 x64
  • CMake 3.17.1
  • Visual Studio 2019
  • OpenCV 4.3.0 with VTK 8.0

---

How to build on Windows

• Open command prompt.
• Run the following commands.

> dir /b
README.md
sources

> mkdir build
> cd build
> cmake ..\sources

• Open the generated .sln file in "build" folder by VisualStudio.
• Build "ALL_BUILD" in Release mode.

---

How to run on Windows

Arguments

• --video="C:\path\to\video"
• --calib="C:\path\to\calib.yml"
• --param="TBD"

> slam.exe --video="C:\path\to\video" --calib="C:\path\to\calib.yml" --param="TBD"

Camera calibration file

• Pinhole camera model is available.
• Please do a calibration using a checkerboards as described in the following page:
  • https://docs.opencv.org/master/d4/d94/tutorial_camera_calibration.html
• Calibration results must be saved like the following:

%YAML 1.1
---
calib:
  numOfPoints: 70
  reprojError: 0.3507153023198588
width: 1280
height: 720
fx: 968.1090867
fy: 979.82648
cx: 637.876864
cy: 354.645136