LoFTR Detector-Free Local Feature Matching with Transformers 环境配置

源码下载链接：https://zju3dv.github.io/loftr

相关模型下载链接: https://pan.baidu.com/s/1yM91Y4JrRiKJjHQ24ubZKw 密码: 2oc9

• 将文件路径切换到解压后的源码文件夹中执行以下命令

# For full pytorch-lightning trainer features (recommended) 过程漫长，需要等待下载···
conda env create -f environment.yaml  
conda activate loftr

# For the LoFTR matcher only
pip install torch einops yacs kornia

cd src/loftr/utils  
wget https://raw.githubusercontent.com/magicleap/SuperGluePretrainedNetwork/master/models/superglue.py 

注意：此处需要修改hosts文件，具体命令如下：

sudo gedit /etc/hosts

#将以下内容添加进hosts文件，点击保存，注意此处的IP地址不一定可以使用，建议去https://google.com.ipaddress.com/drive.google.com查询最新的IP地址
185.199.108.133  raw.githubusercontent.com

/etc/init.d/networking restart

cd demo
./run_demo.sh
#开启摄像头

 python demo_loftr.py --weight ../weights/outdoor_ds.ckpt  --input ../testdata/scannet_test_1500/npu/color --save_video  --save_input

 python demo_loftr.py --weight ../weights/outdoor_ds.ckpt  --input ../testdata/megadepth_test_1500/Undistorted_SfM/0015/images  --save_video --save_input