【MMDetection 系列：一】安装和配置

MMDetection

MMCV

经验

conda create ，不要接着在自己环境里装，可以避免很多坑

安装和配置

micl@ubuntu-gpu ~> conda env list                          
# conda environments:
base                  *  /home/micl/miniconda3
tensorflow-1.4           /home/micl/miniconda3/envs/tensorflow-1.4

根据官方文档
https://github.com/open-mmlab/mmdetection/blob/master/INSTALL.md

Install mmdetection

a. Create a conda virtual environment and activate it. Then install Cython.

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

source activate open-mmlab

conda install cython

b. Install PyTorch stable or nightly and torchvision following the official instructions.

nvcc -V   # 看一下conda 版本

https://pytorch.org/

勾选自己合适的

conda install pytorch torchvision cudatoolkit=9.0 -c pytorch 

另一台机器

conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

c. Clone the mmdetection repository.

git clone https://github.com/open-mmlab/mmdetection.git
cd mmdetection

d. Compile cuda extensions.

./compile.sh

e. Install mmdetection (other dependencies will be installed automatically).

python setup.py develop      # or "pip install -e ."

安装配置成功

其他

建议每次从github中更新时运行步骤e。如果C/CUDA代码有一些更新，也需要运行步骤d。

Prepare COCO dataset.

mmdetection
├── mmdet
├── tools
├── configs
├── data
│   ├── coco
│   │   ├── annotations
│   │   ├── train2017
│   │   ├── val2017
│   │   ├── test2017
│   ├── VOCdevkit
│   │   ├── VOC2007
│   │   ├── VOC2012