Event

LDO v0.1.0 Research Preview Released: Visual Detector and Research Records Open-Sourced

1 reports · 1 sources

Elaybarnoam released v0.1.0 research preview of the visual correspondence detection research tool LDO on GitHub, providing an installable Python 3.11 package and the ldo-research command, integrating MegaLoc and SuperPoint/LightGlue interfaces, and including causal replay, optional stereo/PnP validation, and a 24-second reproducible-on-CPU evidence demo. Usage is limited: model source code and weights are not bundled with the package; learning inference requires bringing your own external source code, checkpoints, and the separate WSL/CUDA configuration described in the installation instructions; the demo shows archived simulated failure evidence rather than real-time inference. Regarding licensing, the original code is Apache-2.0, ORB-SLAM3-related parts use GPL-3.0-or-later, and models, datasets, and image-derived content retain their respective upstream terms.