install.md

Requirements

• Linux or Windows(Experimental)
• Python 3.6+
• PyTorch 1.3 or higher
• mmcv

Installation

a. Create a conda virtual environment and activate it.

conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab

b. Install PyTorch and torchvision following the official instructions. Here we use PyTorch 1.6.0 and CUDA 10.1. You may also switch to other version by specifying the version number.

conda install pytorch=1.6.0 torchvision cudatoolkit=10.1 -c pytorch

c. Install MMCV following the official instructions. Either mmcv or mmcv-full is compatible with MMSegmentation, but for methods like CCNet and PSANet, CUDA ops in mmcv-full is required.

Install mmcv for Linux:

The pre-build mmcv-full (with PyTorch 1.5 and CUDA 10.1) can be installed by running: (other available versions could be found here)

pip install mmcv-full==1.2.2 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/index.html

Install mmcv for Windows (Experimental):

For Windows, the installation of MMCV requires native C++ compilers, such as cl.exe. Please add the compiler to %PATH%.

A typical path for cl.exe looks like the following if you have Windows SDK and Visual Studio installed on your computer:

C:\Program Files (x86)\Microsoft Visual Studio\2019\Professional\VC\Tools\MSVC\14.26.28801\bin\Hostx86\x64

Or you should download the cl compiler from web and then set up the path.

Then, clone mmcv from github and install mmcv via pip:

git clone https://github.com/open-mmlab/mmcv.git
cd mmcv
pip install -e .

Or simply:

pip install mmcv

Currently, mmcv-full is not supported on Windows.

d. Install SETR.

Instead, if you would like to install SETR in dev mode, run following

git clone https://github.com/fudan-zvg/SETR.git
cd SETR
pip install -e .  # or "python setup.py develop"
pip install -r requirements/optional.txt

Note:

1. When training or testing models on Windows, please ensure that all the '\' in paths are replaced with '/'. Add .replace('\', '/') to your python code wherever path strings occur.
2. The version+git_hash will also be saved in trained models meta, e.g. 0.5.0+c415a2e.
3. When SETR is installed on dev mode, any local modifications made to the code will take effect without the need to reinstall it.
4. If you would like to use opencv-python-headless instead of opencv-python, you can install it before installing MMCV.
5. Some dependencies are optional. Simply running pip install -e . will only install the minimum runtime requirements. To use optional dependencies like cityscapessripts either install them manually with pip install -r requirements/optional.txt or specify desired extras when calling pip (e.g. pip install -e .[optional]). Valid keys for the extras field are: all, tests, build, and optional.

A from-scratch setup script

Linux

Here is a full script for setting up SETR with conda and link the dataset path (supposing that your dataset path is $DATA_ROOT).

conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab

conda install pytorch=1.6.0 torchvision cudatoolkit=10.1 -c pytorch
pip install mmcv-full==1.2.2 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/index.html
git clone https://github.com/fudan-zvg/SETR.git
cd SETR
pip install -e .  # or "python setup.py develop"
pip install -r requirements/optional.txt

mkdir data
ln -s $DATA_ROOT data

Windows(Experimental)

Here is a full script for setting up SETR with conda and link the dataset path (supposing that your dataset path is %DATA_ROOT%. Notice: It must be an absolute path).

conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab

conda install pytorch=1.6.0 torchvision cudatoolkit=10.1 -c pytorch
set PATH=full\path\to\your\cpp\compiler;%PATH%
pip install mmcv

git clone https://github.com/fudan-zvg/SETR.git
cd SETR
pip install -e .  # or "python setup.py develop"
pip install -r requirements/optional.txt

mklink /D data %DATA_ROOT%