Oriel — AI & technology worth noticing, every day.

Show HN: DRS-VPT. Localize a photo in a 3D point cloud without an initial pose

Repository · https://github.com/ori-drs/drs-vpt

First seen by Oriel on .

AI summary

DRS-VPT is an open-source Vision Point Transformer that localizes camera images in 3-D point-cloud maps and calibrates camera-LiDAR systems without targets or pose priors.

  • One model supports driving LiDAR scans and indoor maps, with browser demos and Python APIs for localization, intrinsic prediction, and camera-LiDAR calibration.
  • Released evaluations report 99.71% accuracy on KITTI and 97.62% on nuScenes, using recall within 2 meters and 5 degrees.
  • Map encoding can be reused across image streams, making subsequent localization pay only for image processing; joint calibration accepts multiple synchronized image-scan pairs.
  • Installation supports NVIDIA GPU, CPU-only Linux, and macOS, while model weights require a 2.9 GB download and the first macOS run compiles spconv.
  • The Apache 2.0 code accompanies model weights licensed CC BY-NC 4.0 for non-commercial use only.

#3d localization #point clouds #camera-lidar calibration #vision transformers

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hacker News: 9, latest observation ; +6 over observed period since

Comments

Comment count reported by the source.

  • Hacker News: 0, latest observation ; 0 over observed period since

Aliases

SourceExternal IDURL
Hacker News50472560https://github.com/ori-drs/drs-vpt Read discussion