kimsoohwan / fastcampus_slam_codes

Code exercises for the SLAM course in 'Computer Vision, LiDAR processing, and Sensor Fusion for Autonomous Driving' lecture series

fastcampus_slam_codes

This repository contains code exercises for the SLAM section in the lecture series - 'Computer Vision, LiDAR processing, and Sensor Fusion for Autonomous Driving' at FastCampus. This lecture series is delivered in Korean language.

How to use

Most of the code exercises are based on the base docker image. The base docker image contains numerous C++ libraries for SLAM, such as OpenCV, Eigen, Sophus, PCL, and ceres-solver.

You can build the base docker image using the following command.

docker build . --tag slam:latest --progress=plain

Table of contents

• Chapter 1: Introduction to SLAM
  • 1.1 Lecture introduction
  • 1.2 What is SLAM?
  • 1.3 Hardware for SLAM
  • 1.4 Types of SLAM
  • 1.5 Applications of SLAM
  • 1.6 Before we begin...
  • 1.7 Basic C++ / CMake
• Chapter 2: Introduction 3D Spaces
  • 2.1 3D rotation and translation
  • 2.2 3D rotation and translation, using Eigen library
  • 2.3 Homogeneous coordinates
  • 2.4 Lie Group
  • 2.5 Basic Lie algebra
  • 2.6 Lie Group and Lie algebra, using Sophus library
  • 2.7 How cameras work
  • 2.8 How LiDARs work
• Chapter 3: Image processing
  • 3.1 Local feature extraction & matching
  • 3.2 Local feature extraction & matching, using OpenCV library
  • 3.3 Superpoint and Superglue, using C++ and TensorRT
  • 3.4 Global feature extraction
  • 3.5 Bag of Visual Words, using DBoW2 library
  • 3.6 Learning-based global feature extraction, using PyTorch and Tensorflow libraries
  • 3.7 Feature tracking
  • 3.8 Optical flow, using OpenCV library
• Chapter 4: Point cloud processing
  • 4.1 Introduction to point cloud processing
  • 4.2 Introduction to point cloud processing, using PCL library
  • 4.3 Point cloud pre-processing
  • 4.4 Point cloud pre-processing, using PCL library
  • 4.5 Iterative closest point
  • 4.6 Iterative closest point, using PCL library
  • 4.7 Advanced ICP methods
  • 4.8 Advanced ICP methods (G-ICP, NDT, TEASER++, KISS-ICP), using PCL library
  • 4.9 Octree, Octomap, Bonxai, using PCL/Octomap/Bonxai libraries
• Chapter 5: Multiple view geometry
  • 5.1 Epipolar geometry
  • 5.2 Essential and Fundamental matrix estimation, using OpenCV library
  • 5.3 Homography
  • 5.4 Bird's eye view (BEV) projection, using OpenCV library
  • 5.5 Simple monocular visual odometry, using OpenCV library
  • 5.6 Triangulation
  • 5.7 Triangulation, using OpenCV library
  • 5.8 Perspective-n-Points (PnP) and Direct Linear Transform (DLT)
  • 5.9 Fiducial marker tracking, using OpenCV library
  • 5.10 RANSAC
  • 5.11 Advanced RANSAC methods (USAC)
  • 5.12 RANSAC and USAC, using OpenCV and RansacLib libraries
  • 5.13 Graph-based SLAM
  • 5.14 Least squares
  • 5.15 Schur complement
  • 5.16 Bundle adjustment
  • 5.17 Bundle adjustment, using Ceres-Solver library
• Chapter 6: Visual-SLAM
  • 6.1 Overview of feature-based VSLAM
  • 6.2 Overview of direct VSLAM
  • 6.3 Overview of visual-inertial odometry (VIO)
  • 6.4 Spatial AI
  • 6.5 ORB-SLAM2, ORB-SLAM3
  • 6.6 DynaVINS
  • 6.7 CubeSLAM
• Chapter 7: LiDAR SLAM
  • 7.1 Overview of 2D LiDAR SLAM
  • 7.2 Overview of 3D LiDAR SLAM and LiDAR-inertial odometry
  • 7.3 HDL-Graph-SLAM
  • 7.4 KISS-ICP
  • 7.5 SHINE-Mapping
• Chapter 8: CI/CD for SLAM
  • 8.1 TDD and tests
  • 8.2 CI/CD
  • 8.3 CI agents
  • 8.4 CI/CD for Python SLAM projects
  • 8.5 CI/CD for C++ SLAM projects
• Final projects:
  • DSP-SLAM
  • Suma++

Acknowledgements

ORB-SLAM 2/3 authors, DynaVINS authors, CubeSLAM authors, HDL-Graph-SLAM authors, KISS-ICP authors, SHINE-Mapping authors, and all the authors of the libraries used in this repository.