-
The library requires PyTorch~=1.13 and MinkowskiEngine library to be installed manually. See PyTorch website and MinkowskiEngine repository for the detailed instructions.
-
Another option is to use the suggested Dockerfile. The following commands should be used to build, start and enter the container:
-
Build the image
bash docker/build.sh
-
Start the container with the datasets directory mounted:
bash docker/start.sh [DATASETS_DIR]
-
Enter the container (if needed):
bash docker/into.sh
-
-
After the pre-requisites are met, install the Open Place Recognition library with the following command:
pip install .
Currently only MinkLoc++ pretrained on Oxford RobotCar available. You can download it using google drive link (the link is subject to change).
If everything is installed correctly, you can use the library like below:
from opr.models import minkloc_multimodal
baseline_model = minkloc_multimodal(weights="path_to_checkpoint")
MIT License (the license is subject to change in future versions)