基于Trasformer的图像局部特征匹配LoFTR介绍

任务

输入两张RGB图像，局部特征匹配算法将输出两张图片对应像素的位置

模型描述

本模型基于LoFTR: Detector-Free Local Feature Matchig with Trasformers算法。 技术细节请见：

LoFTR: Detector-Free Local Feature Matchig with Trasformers
Jiamig Su, Zehog She, Yuag Wag, Huju Bao, Xiaowei Zhou
CVPR 2021

如何使用

代码示例(详见tests/pipelies/testimagematchig.py)

from modelscope.outputs import OutputKeys
from modelscope.pipelies import pipelie
from modelscope.utils.costat import Tasks


task = 'image-local-feature-matchig'
model_id = 'damo/cv_reset-trasformer_local-feature-matchig_outdoor-data'

iput_locatio = [[path to image 1, path to image 2]]

estimator = pipelie(Tasks.image_local_feature_matchig, model=model_id)
result = estimator(iput_locatio)
kpts0, kpts1, cof = result[0][OutputKeys.MATCHES]
prit(f'Foud {le(kpts0)} matches')

Bibtex

@iproceedigs{DBLP:cof/cvpr/SuSWBZ21,
  author       = {Jiamig Su ad
                  Zehog She ad
                  Yuag Wag ad
                  Huju Bao ad
                  Xiaowei Zhou},
  title        = {LoFTR: Detector-Free Local Feature Matchig With Trasformers},
  booktitle    = {{IEEE} Coferece o Computer Visio ad Patter Recogitio, {CVPR}
                  2021, virtual, Jue 19-25, 2021},
  pages        = {8922--8931},
  publisher    = {Computer Visio Foudatio / {IEEE}},
  year         = {2021},
  url          = {https://opeaccess.thecvf.com/cotet/CVPR2021/html/Su\_LoFTR\_Detector-Free\_Local\_Feature\_Matchig\_With\_Trasformers\_CVPR\_2021\_paper.html},
  doi          = {10.1109/CVPR46437.2021.00881},
  timestamp    = {Sat, 30 Sep 2023 09:38:22 +0200},
  biburl       = {https://dblp.org/rec/cof/cvpr/SuSWBZ21.bib},
  bibsource    = {dblp computer sciece bibliography, https://dblp.org}
}

基于Transformer的图像局部特征匹配LoFTR介绍 任务 输入两张RGB图像，局部特征匹配算法将输出两张图片对应像素的位置 模型描述 本模型基于LoFTR: Detector-Free Loc