install_commands.txt

OUTDATED PARTIALLY (customized repo is in use now)

git clone https://github.com/jonasschult/mask3d
export TORCH_CUDA_ARCH_LIST="6.0 6.1 6.2 7.0 7.2 7.5 8.0 8.6"
cd mask3d/
vim environment.yaml (comment pyyaml and pycocotools, change conda name to thaumato)
conda env create -f environment.yml
conda activate thaumato
conda install -c "nvidia/label/cuda-11.3.1" cuda-toolkit
pip3 install torch==1.12.1+cu113 torchvision==0.13.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
pip3 install torch-scatter -f https://data.pyg.org/whl/torch-1.12.1+cu113.html
pip3 install 'git+https://github.com/facebookresearch/detectron2.git@710e7795d0eeadf9def0e7ef957eea13532e34cf' --no-deps
cd third_party/
git clone --recursive "https://github.com/NVIDIA/MinkowskiEngine"
cd MinkowskiEngine
git checkout 02fc608bea4c0549b0a7b00ca1bf15dee4a0b228
python setup.py install --force_cuda --blas=openblas
cd ..
git clone https://github.com/ScanNet/ScanNet.git
cd ScanNet/Segmentator
git checkout 3e5726500896748521a6ceb81271b0f5b2c0e7d2
make
cd ../../pointnet2
python setup.py install
cd ../../
pip3 install pytorch-lightning==1.7.2
pip3 install torchmetrics==0.11.4

TRAINING COMMANDS:

generate training samples with 
syntethic data:
generate_surface_pointgroup_dataset.py
and actual data:
generate_training_samples_from_annotations.py

place them in data/raw/thaumatoanakalyptor test, train, validation.

running:
python3 -m datasets.preprocessing.thaumato_preprocessing preprocess --data_dir="data/raw/thaumatoanakalyptor" --save_dir="data/processed/stpls3d"

generate ThaumatoAnakalyptor/mask3d/data/processed/stpls3d/train_database.yaml properly:
mask3d/conf/data/datasets/stpls3d.yaml has either ThaumatoAnakalyptor/mask3d/ before for example conf/augmentation/albumentations_aug.yaml (which certainly works during inference with the prepend. or not (with works for the training). maybe the inference works as well let's see)

source ./scripts/stpls3d/stpls3d_train.sh

test:
source ./scripts/stpls3d/stpls3d_test.sh