Qualcomm® AI HubAI Hub

LightGlue

Lightweight attention‑based matcher for sparse local image features.

LightGlue is a deep neural network that matches sparse local features across image pairs. It replaces the traditional combination of nearest‑neighbour matching and robust geometric fitting with a fast, lightweight attention‑based model, performing both matching and a learned pruning of non‑matchable points in a single forward pass. LightGlue consumes keypoints and descriptors already produced by a separate feature extractor (SuperPoint) and outputs point correspondences between the two images. This recipe pairs LightGlue's official checkpoint (trained against Magic‑Leap's SuperPoint) with this repo's MIT‑licensed SuperPoint port, so reported accuracy is directionally meaningful but not a direct reproduction of the paper's figures.

Not supported

This model is currently not supported on any Automotive chipset.

To see performance metrics for this model on other chipsets, click the button below.

View for other chipsets

Technical Details

Descriptor dimension:256
Input resolution (pixel coords):480x640
Max input keypoints per image:500
Model checkpoint:superpoint_lightglue_v0-1_arxiv.pth
Model size (float):43.7 MB
Number of parameters:11.3M
Number of transformer layers:9

Applicable Scenarios

  • Augmented Reality
  • Visual Localisation
  • Robotics
  • Autonomous Vehicles

License

Tags

  • real-time

Supported Automotive Devices

  • SA8255P ADP
  • SA8295P ADP
  • SA8650P ADP
  • SA8775P ADP

Supported Automotive Chipsets

  • Qualcomm® SA8255P
  • Qualcomm® SA8295P
  • Qualcomm® SA8650P
  • Qualcomm® SA8775P

Related Models

See all models

Looking for more? See models created by industry leaders.

Discover Model Makers