zfd1/hift

dense feature descriptor for image matching

Learning Enriched Feature Descriptor for Image Matching and Visual Measurement

demo and pre-trained weight of HIFT --- a dense descriptor for local feature matching.

Our work was accepted by IEEE Transactions on Instrumentation & Measurement 2023, and can be accessed via manuscript.

Pre-Trained Weights

We trained our HIFT with one-stage end-to-end triplet training strategy on MS-COCO, Multi-illumination and VIDIT datasets (same as LISRD) and the pre-trained weight is available at hift

Model file

The core implementation of HIFT is shown in HIFT_core.py

DEMO：SIFT+HIFT

1. We provide the image matching demo of using SIFT keypoints and HIFT descriptor in demo.ipynb.
2. We provide the demo of exporting SIFT keypoints and HIFT descriptor in export_descriptor_sift.py, and it can be easily modified to other off-the-shelf detectors and matchers for evaluation.

CUDA_VISIBLE_DEVICES=0 python export_descriptor_sift.py

For more evaluation details, please refer to the LISRD

Citation

If you are interested in this work, please cite the following work:

@ARTICLE{10058693,
  author={Rao, Yuan and Ju, Yakun and Wang, Sen and Gao, Feng and Fan, Hao and Dong, Junyu},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={Learning Enriched Feature Descriptor for Image Matching and Visual Measurement}, 
  year={2023},
  volume={72},
  number={},
  pages={1-12},
  doi={10.1109/TIM.2023.3249237}}

Acknowledgments

Our work is based on LISRD and we use their code. We appreciate the previous open-source repository LISRD