finds.dev← search

// the find

gaoxiang12/slambook2

★ 6,727 · C++ · MIT · updated Dec 2024

edition 2 of the slambook

This is the companion code for Gao Xiang's "14 Lectures on Visual SLAM" (second edition), the most widely used Chinese-language SLAM textbook. Each chapter folder (ch2–ch13) is a self-contained exercise walking from Eigen/Lie-group math up to a working monocular/stereo visual odometry pipeline with bundle adjustment and loop closure. It's for someone learning SLAM fundamentals hands-on, not for pulling in a production SLAM library.

Chapters build in a sensible progression — basic transforms, then epipolar geometry and PnP/ICP, then optical flow and direct methods, then a full frontend/backend/viewer VO system in ch13's myslam module with threads split the way a real system would be. It uses actual industry tools instead of toy reimplementations: Ceres and g2o for bundle adjustment, Sophus for SO(3)/SE(3) math, DBoW3 for loop closure, Pangolin for 3D visualization. Because it's the de facto standard teaching reference in the Chinese SLAM community, build problems and gotchas are already extensively documented across forums and issues.

The README gives no build instructions, dependency versions, or chapter overview in English — without the book itself (sold separately, mostly untranslated) the repo is close to opaque. Compiled artifacts and binaries are checked straight into git (ch2/CMakeFiles, compiled executables like ch2/helloSLAM), which bloats the clone and risks stale binaries shadowing source changes. Dependencies are heavy and unpinned (Ceres, g2o, Pangolin, Sophus, DBoW3 all pulled as submodules or expected to be installed system-wide) with no CI and no version pins, so a fresh environment setup is hours of manual wrangling. Test coverage is a single triangulation test file — this is teaching code with no correctness guarantees to build anything serious on top of.

View on GitHub →

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →