jac99/MinkLoc3Dv2

MinkLoc3Dv2: Improving Point Cloud Based Place Recognition with Ranking-based Loss and Large Batch Training

43
/ 100
Emerging

This project helps autonomous vehicles, robots, and mapping systems recognize where they are by matching current 3D point cloud scans to a database of known locations. It takes raw 3D point cloud data as input and outputs a highly discriminative descriptor that accurately identifies the location. Robotics engineers, autonomous driving researchers, and anyone developing spatial recognition systems would use this to improve their system's ability to localize itself in complex environments.

No commits in the last 6 months.

Use this if you need to build or enhance a system for robust 3D place recognition using point cloud data, especially in large-scale or challenging environments.

Not ideal if your application relies on 2D image data for localization, or if computational efficiency on highly resource-constrained edge devices is your absolute top priority without specialized hardware.

autonomous-navigation robotics-localization 3d-mapping place-recognition lidar-processing
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

94

Forks

17

Language

Python

License

MIT

Last pushed

Jan 31, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/jac99/MinkLoc3Dv2"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.