// the find
rust-cv/cv
Rust CV mono-repo. Contains pure-Rust dependencies which attempt to encapsulate the capability of OpenCV, OpenMVG, and vSLAM frameworks in a cohesive set of APIs.
A Rust mono-repo attempting to replicate the photogrammetry and SLAM capabilities of OpenCV/OpenMVG without any C++ dependencies. The core algorithms are there — AKAZE feature extraction, essential matrix estimation, P3P pose solving, ARRSAC consensus — but the overall system stops well short of a usable SfM or SLAM pipeline. For researchers or Rust developers who need individual low-level CV primitives, the individual crates are the real value here.
AKAZE implementation is solid and `#[no_std]`-compatible, which is genuinely useful for embedded targets where OpenCV is a non-starter. The `sample-consensus` abstraction is clean — ARRSAC, estimators, and models are properly separated so you can plug in your own pieces. Lambda Twist P3P has mathematical derivation notes alongside the code, which is rare and makes it auditable. The mono-repo consolidation means the crates actually build together and dependency versions stay consistent.
The README's own checklist is damning: loop closure, reconstruction, meshing, densification, and homography estimation are all unchecked. This is a collection of algorithm building blocks, not a working SfM system — you cannot hand it a folder of images and get a point cloud. Last meaningful SfM activity (cv-sfm) appears stalled; the tutorial stops at chapter 4 of what should be a much longer series. SIFT and ORB are missing, which means you're stuck with AKAZE for feature extraction — fine for many cases but a real gap if you need compatibility with existing pipelines or datasets that expect SIFT descriptors.