Repository for Deep Learning based petrography of igneous Plagioclase crystals based on circular polarised light images of thin sections. We make extensive use of the MMDetection library, with the work based on DetectoRS models.
In order to install this private package you must be able to access it (which you can if you're reading this) and run have/create a python 3.7 environment for relevant package requirements (PyTorch can be a pain like that).
Ensure GCC is installed on your system.
Create and activate environment:
eg conda create -n MinDetEnv python=3.7
eg conda activate MinDetEnv
Install required libraries (cluster nodes):
wget https://download.pytorch.org/whl/cu110/torch-1.7.0%2Bcu110-cp37-cp37m-linux_x86_64.whl
pip install torch-1.7.0+cu110-cp37-cp37m-linux_x86_64.whl
pip install torchvision==0.8.0 torchaudio==0.7.0
pip install openmim
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7.0/index.html
mim install "mmdet<3.0.0"
Install required libraries (non-cluster):
pip install torch==1.7.0+cu110 torchvision==0.8.0 torchaudio==0.7.0 -f https://download.pytorch.org/whl/torch_stable.html
OR pip install torch==1.7.0 torchvision==0.8.0 torchaudio==0.7.0 (mac users)
pip install openmim
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7.0/index.html
OR pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.7.0/index.html (mac users)
mim install "mmdet<3.0.0"
Install using the following command:
pip install git+https://git@github.com/norberttoth398/MinDet
Please cite the following publication describing the present software:
Toth, N. and Maclennan, J. (2024) “MinDet1: A deep learning-enabled approach for plagioclase textural studies”, Volcanica, 7(1), pp. 135–151. doi: 10.30909/vol.07.01.135151.