SpatialHub/efficient-loftr-onnx

EfficientLoFTR ONNX Weights

This repository contains the ONNX-optimized weights for EfficientLoFTR, a model used for finding matching points between pairs of images.

By converting the original PyTorch model weights into the ONNX format, these files allow you to run fast feature-matching inference on both CPU and GPU without needing to install the heavy PyTorch framework.

Available Files

• —`eloftr_outdoor_full.onnx`: The standard version of the model, optimized for the best matching quality.
• —`eloftr_outdoor_opt.onnx`: An efficiency-focused version of the model, optimized for faster inference speed.

How to Use

The easiest way to load and use these files is through the [spatialhub](https://github.com/pankajkaushik12/spatialhub) Python library.

Original Citation

If you use these models in academic work, please cite the original authors:

@inproceedings{wang2022efficientloftr,
  title={EfficientLoFTR: Semi-Dense Local Feature Matching with Sparse Transformers},
  author={Wang, Yanzhao and Geng, Yuwei and Jiang, Zheng and Zhao, Yihong and Jin, Shisheng and Lin, Siyu and Han, Feng},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2022}
}

Note: This repository provides pre-converted weights for inference purposes.