// the find
gaoxiang12/slambook
Companion code for the book '14 Lectures on Visual SLAM' by Gao Xiang. Walks through the full classical SLAM pipeline chapter by chapter: Eigen, Sophus, OpenCV, Ceres, g2o, feature-based and direct odometry, loop closure with DBoW3, dense reconstruction. Aimed at students and robotics engineers learning SLAM from scratch.
The chapter-by-chapter structure maps directly to the book's theory, so you're never guessing why a piece of code exists. Covers both optimization backends side by side — Ceres and g2o — for the same problems, which is genuinely useful for understanding the tradeoffs. The project directory shows an incremental VO system evolving from 0.1 to 0.4, making the design decisions visible rather than presenting a finished black box. Includes real dataset files and pre-built vocabularies so examples actually run without hunting down data.
This is v1, effectively abandoned since the author released slambook2 in 2019 — Ubuntu 14/16 era dependencies, pinned to older g2o and Sophus APIs that have breaking changes in current versions. The build system is a pile of per-chapter CMakeLists with copy-pasted FindXXX modules rather than a unified build, so getting everything to compile on a modern system is a weekend project. No tests anywhere — the code is read-and-run-once pedagogy, not something you'd fork and maintain. Dense reconstruction chapter depends on REMODE which is essentially unmaintained and GPU-only.