Currently, most 3D vision algorithms are written in C++, such as the SLAM algorithm. Since these algorithms are all written in C++, there must be reasons for this. The main reason is that the SLAM system has very high real-time requirements, and both speed and accuracy are crucial. Python also has acceleration methods, but overall, C++ is closer to the hardware level, resulting in much better execution efficiency and resource usage. Moreover, many languages are encapsulated based on C++ or C.
C++ has always been regarded as the most difficult programming language. Having worked in this field for many years, I also believe this to be true. Compared to several other languages, C++ is relatively difficult purely from a syntax structure perspective. Furthermore, many students do not have a computer science background and their foundation is not very solid. The SLAM algorithm integrates many modules from C++ and includes a large number of third-party libraries, making it even more challenging for beginners to understand the source code of these algorithms. However, these source codes are not without clues, as many algorithms follow relatively fixed patterns.
The following images show part of the source code from a certain SLAM algorithm.
In this course, we will first spend a few lectures covering the basics of C++, then focus on several commonly used C++ modules in 3D vision, mainly including pointers, references, classes and objects, macros and templates, STL programming, stacks and queues, smart pointers, and CMakeLists. Additionally, we will provide plenty of practical examples and discuss interview skills and questions during the course. Our goal is to thoroughly understand some of the more challenging modules in C++ and master the writing patterns of algorithms like SLAM, as well as the interview skills for C++ related positions.
