- Linux
- Python 3.6+
- PyTorch 1.3+
- CUDA 9.2+
- GCC 5+
- mmcv 0.6.2
a. Create a conda virtual environment and activate it.
conda create -n env_name python=3.7 -y
conda activate env_name
b. Install PyTorch and torchvision following the official instructions, e.g.,
conda install pytorch torchvision -c pytorch
Note: Make sure that your compilation CUDA version and runtime CUDA version match. You can check the supported CUDA version for precompiled packages on the PyTorch website.
e.g.
If you have CUDA 10.1 installed under /usr/local/cuda
and would like to install
PyTorch 1.6.0, you need to install the prebuilt PyTorch with CUDA 10.1.
conda install pytorch=1.6.0 cudatoolkit=10.1 torchvision=0.7.0 -c pytorch
e.g.
If you have CUDA 9.2 installed under /usr/local/cuda
and would like to install
PyTorch 1.3.1., you need to install the prebuilt PyTorch with CUDA 9.2.
conda install pytorch=1.3.1 cudatoolkit=9.2 torchvision=0.4.2 -c pytorch
c. Clone the EDA repository.
git clone https://github.com/ShichengMiao16/EDA.git --recursive
cd EDA
d. Install build requirements and then install EDA.
# install BboxToolkit
cd BboxToolkit
pip install -v -e . # or "python setup.py develop"
cd ..
# install EDA
pip install -r requirements/build.txt
pip install pycocotools
pip install mmcv==0.6.2
pip install -v -e . # or "python setup.py develop"
Note:
-
The git commit id will be written to the version number with step d. The version will also be saved in trained models. It is recommended that you run step d each time you pull some updates from github. If C++/CUDA codes are modified, then this step is compulsory.
Important: Be sure to remove the
./build
folder if you reinstall mmdet with a different CUDA/PyTorch version.pip uninstall mmdet rm -rf ./build find . -name "*.so" |xargs rm
-
Following the above instructions, EDA is installed on
dev
mode, any local modifications made to the code will take effect without the need to reinstall it (unless you submit some commits and want to update the version number). -
If you would like to use
opencv-python-headless
instead ofopencv-python
, you can install it before installing MMCV. -
Some dependencies are optional. Simply running
pip install -v -e .
will only install the minimum runtime requirements. To use optional dependencies likealbumentations
andimagecorruptions
either install them manually withpip install -r requirements/optional.txt
or specify desired extras when callingpip
(e.g.pip install -v -e .[optional]
). Valid keys for the extras field are:all
,tests
,build
, andoptional
. -
If you install torchvision < 0.5.0, please run
pip install "pillow<9"
to avoidImportError
.
The code can be built for CPU only environment (where CUDA isn't available).
In CPU mode you can run the $EDA/demo/webcam_demo.py
for example.
However some functionality is gone in this mode:
- Deformable Convolution
- Deformable ROI pooling
- CARAFE: Content-Aware ReAssembly of FEatures
- nms_cuda
- sigmoid_focal_loss_cuda
So if you try to run inference with a model containing deformable convolution, you will get an error.
Note: We set use_torchvision=True
on-the-fly in CPU mode for RoIPool
and RoIAlign
We provide a Dockerfile to build an image. Ensure that your docker version >= 19.03.
# build an image with PyTorch 1.6.0 and CUDA 10.1
# If you prefer other versions, just modify the Dockerfile
docker build -t eda $EDA/docker/
Run it with
docker run --gpus all --shm-size=8g -it -v ${DATA_DIR}:/EDA/data eda