Skip to content

LoFTR implemented with mindspore framework

Notifications You must be signed in to change notification settings

VictorHe-1/LoFRT-ms

 
 

Repository files navigation

LoFTR-ms

LoFTR implemented using MindSpore framework

Preprocessing Dataset

  1. Convert NPZ files for training:
python scripts/mega_depth_idx_convert.py 
  1. Convert image files for training:
python scripts/mega_depth_preprocess.py 
  1. Convert NPZ files for testing:
python scripts/mega_depth_test_idx_convert.py 
  1. Create link:
sh scripts/mega_depth_link.sh 

Training

bash scripts/reproduce_train/outdoor_ds.sh 

Distribute Training

Before performing distributed training, it is necessary to modify the loftr configuration file as follows:

- cfg.system.distribute = False
+ cfg.system.distribute = True

Then run the following command:

bash scripts/reproduce_train/distribute_outdoor_ds.sh 

Testing

bash scripts/reproduce_test/outdoor_ds.sh 

About

LoFTR implemented with mindspore framework

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 98.9%
  • Shell 1.1%