# cvg/lightglue

**Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/cvg-lightglue).**

_How this analysis was created: the description and tags below were written by an AI model that read this project's README and public documentation pages; stars, license and language come straight from the GitHub API. The model does not read the source code._

4,625 stars · 539 forks · Python · Apache-2.0

## Links

- GitHub: https://github.com/cvg/LightGlue
- awesome-repositories: https://awesome-repositories.com/repository/cvg-lightglue.md

## Description

LightGlue is a deep learning framework designed for local feature matching and high-speed correspondence estimation between pairs of images. It functions as a computer vision matching model that identifies corresponding keypoints across different viewpoints.

The system utilizes an adaptive neural network architecture that dynamically optimizes inference speed by pruning its own depth and width based on the input image pairs. This approach employs a transformer-style attention mechanism and cross-image attention to compute correlations between feature descriptors.

The matching process includes an iterative refinement loop and dynamic early stopping to halt computation once confidence thresholds are met. These capabilities support a broader computer vision pipeline for real-time image alignment and neural network inference optimization.

## Tags

### Artificial Intelligence & ML

- [Dynamic Architecture Pruning](https://awesome-repositories.com/f/artificial-intelligence-ml/adaptive-parameter-prunings/dynamic-architecture-pruning.md) — Implements a neural network that dynamically prunes its own depth and width to optimize inference speed.
- [Computer Vision Models](https://awesome-repositories.com/f/artificial-intelligence-ml/computer-vision-models.md) — Provides a deep learning model designed for high-speed correspondence estimation in image pairs.
- [Inference Performance Optimization](https://awesome-repositories.com/f/artificial-intelligence-ml/inference-performance-optimization.md) — Optimizes inference performance by dynamically adjusting network complexity based on input image pairs. ([source](https://github.com/cvg/lightglue#readme))
- [Keypoint Descriptor Extractors](https://awesome-repositories.com/f/artificial-intelligence-ml/keypoint-detection/keypoint-descriptor-extractors.md) — Represents image content as a sparse set of local descriptors extracted from keypoints.
- [Cross-Attention Mechanisms](https://awesome-repositories.com/f/artificial-intelligence-ml/multi-modal-tokenizers/multi-modal-embedding-models/cross-attention-mechanisms.md) — Employs cross-attention mechanisms to compute correlations between feature descriptors of two different images.
- [Adaptive Neural Networks](https://awesome-repositories.com/f/artificial-intelligence-ml/neural-network-architectures/adaptive-neural-networks.md) — Implements an architecture that dynamically prunes its own depth and width to optimize inference speed.
- [Inference Speed Optimizers](https://awesome-repositories.com/f/artificial-intelligence-ml/token-optimization-utilities/generation-speed-optimizers/transcription-speed-optimizers/inference-speed-optimizers.md) — Reduces computational cost and increases processing speed through adaptive network pruning during inference.
- [Correspondence Refinement Loops](https://awesome-repositories.com/f/artificial-intelligence-ml/agentic-systems-frameworks/agentic-workflows/iterative-refinement-workflows/mask-refinement-loops/correspondence-refinement-loops.md) — Refines matching assignments over multiple passes to increase the precision of point correspondences.
- [Computer Vision Pipelines](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/frameworks/computer-vision/computer-vision-pipelines.md) — Processes image data to identify shared landmarks for use in broader computer vision workflows.
- [Inference Early Stopping](https://awesome-repositories.com/f/artificial-intelligence-ml/model-training/early-stopping-callbacks/inference-early-stopping.md) — Halts the iterative matching process once the network reaches a predefined confidence threshold.

### Part of an Awesome List

- [Deep Feature Matchers](https://awesome-repositories.com/f/awesome-lists/ai/deep-learning-and-computer-vision/deep-feature-extractors/deep-feature-matchers.md) — Utilizes a neural network to identify corresponding keypoints between image pairs for computer vision tasks.
- [Attention-Based Matching](https://awesome-repositories.com/f/awesome-lists/devtools/local-feature-matching/attention-based-matching.md) — Uses a transformer-style attention mechanism to compute correlations and match feature descriptors.
- [Local Feature Matching](https://awesome-repositories.com/f/awesome-lists/devtools/local-feature-matching.md) — High-speed local feature matching using transformer-based architectures.

### Graphics & Multimedia

- [Real-Time Image Alignments](https://awesome-repositories.com/f/graphics-multimedia/real-time-image-alignments.md) — Matches visual features quickly between frames to enable low-latency tracking and image alignment.