- Linux
- Python 3.5/3.6/3.7
- PyTorch 1.1/1.3.1
- CUDA 10.0/10.1
- NCCL 2+
- GCC 4.9+
- mmcv<=0.2.14
We have test the following environment:
Python 3.7
Pytorch 1.1.0
CUDA 10.0
mmcv==0.2.13
a. Create a conda virtual environment and activate it. Then install Cython.
conda create -n boxlevelset python=3.7 -y
source activate boxlevelset
conda install cython
b. Install PyTorch and torchvision following the official instructions.
conda install pytorch=1.3.1 torchvision cudatoolkit=10.0 -c pytorch -y
Note:
- If you want to use Pytorch>1.5, you have to made some modifications to the
cuda ops
. See here for a reference. - There is a known bug happened to some users but not all (As I have successfully run it on V100 and Titan Xp). If it occurs, please refer to here.
c. Clone the boxlevelset repository.
https://github.com/LiWentomng/boxlevelset.git
cd boxlevelset
d. Compile cuda extensions.
bash compile.sh
e. Install boxlevelset (other dependencies will be installed automatically).
python setup.py develop
# or "pip install -e ."
sudo apt-get install swig
cd DOTA_devkit
swig -c++ -python polyiou.i
python setup.py build_ext --inplace