// the take
A single-header C++ implementation of quadric error metric mesh decimation - it collapses edges to hit a target triangle count. Aimed at graphics/game/3D-tooling developers who need to cut polycount on watertight meshes without pulling in Meshlab or a full geometry-processing library.
Core logic is one header (Simplify.h) at ~400 lines with no dependencies beyond the STL, so it's trivial to vendor into an existing codebase. The threshold-based collapse (skip sorting, just collapse under a cost cutoff) is the actual reason it beats Meshlab on speed, not just a marketing claim. It's been ported into enough serious consumers (Unity, .NET's MeshDecimator, 3D Slicer's medical imaging toolchain, several WASM builds) that the core algorithm has had real-world exposure well beyond the original author's testing.
The README admits upfront it breaks down on non-watertight meshes or thin sections, which rules out a lot of real-world input (scans, CAD exports with open boundaries) without a repair pass first. There's no test suite in the repo itself - confidence comes from third-party ports, not from anything the maintainer ships. The repo is cluttered with stale, unrelated binaries and dependencies (bin32/bin64 executables, GLEW, DevIL) that have nothing to do with the actual simplification code and just add noise when you're trying to find what to vendor. Documentation stops at "call simplify_mesh(target_count)" - no notes on degenerate-triangle handling, numerical stability at high reduction ratios, or thread-safety.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →