- 数据集配置文件: idrid.py , ddr.py
- 模型配置文件示例: config_sample.py
- 一个简单的调试文件: debug.py
## CUDA_VISIBLE_DEVICES表示使用哪些序号的显卡
## PORT可随意设置,不冲突就行,为程序使用的端口(传数据、syncbn等)
## 最后的数字表示启用几个线程,用n张卡就等于n
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=23456 ./tools/dist_train.sh ./configs/_idrid_/fcn_hr48_40k_idrid_bdice.py 4
# or
CUDA_VISIBLE_DEVICES=2,3 PORT=23456 ./tools/dist_train.sh ./configs/_idrid_/debug.py 2
# 从checkpoint处继续训练
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=23456 ./tools/dist_train.sh ./configs/_idrid_/fcn_hr48_40k_idrid_bdice.py 4 --resume-from ./work_dirs/xxx/xxx.pth
# 以下两种结果应该是一样的
# 多卡
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=23456 ./tools/dist_test.sh ./configs/xxx.py ./work_dirs/xxx/xxx.pth 4 --eval-mIoU
# 单卡
CUDA_VISIBLE_DEVICES=2 python tools/test.py ./configs/xxx.py ./work_dirs/xxx/xxx.pth --eval mIoU
基本环境:
- python=3.7
- torch=1.6.0
- cuda=10.2
- mmcv=1.2.0
# 创建环境 open-mmlab
conda create -n open-mmlab python=3.7 -y
# 激活环境
conda activate open-mmlab
# 之后的命令一定要在这个环境下
conda install pytorch=1.6.0 torchvision cudatoolkit=10.2 -c pytorch -y
pip install mmcv-full==1.2.0 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.6.0/index.html -i https://pypi.douban.com/simple/
pip install opencv-python -i https://pypi.douban.com/simple/
pip install scipy -i https://pypi.douban.com/simple/
pip install tensorboard tensorboardX -i https://pypi.douban.com/simple/
pip install sklearn -i https://pypi.douban.com/simple/
pip install terminaltables -i https://pypi.douban.com/simple/
pip install matplotlib -i https://pypi.douban.com/simple/
# 以下是cuda 10.1版本的环境
conda install pytorch=1.6.0 torchvision cudatoolkit=10.1 -c pytorch -y
pip install mmcv-full==1.2.0 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.6.0/index.html -i https://pypi.douban.com/simple/
pip install opencv-python -i https://pypi.douban.com/simple/
pip install scipy -i https://pypi.douban.com/simple/
pip install tensorboard tensorboardX -i https://pypi.douban.com/simple/
pip install sklearn -i https://pypi.douban.com/simple/
pip install terminaltables -i https://pypi.douban.com/simple/
pip install matplotlib -i https://pypi.douban.com/simple/
cd mmsegmentation-lesion
chmod u+x tools/*
pip install -e . -i https://pypi.douban.com/simple/
(用于对比和迁移到其他版本的mmsegmentation)
- lesion_metrics.py
- lesion_dataset.py
- encoder_decoder_lesion.py
- cascade_encoder_decoder_lesion.py
- binary_loss.py
anaconda安装
wget -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-2020.02-Linux-x86_64.sh
sh Anaconda3-2020.02-Linux-x86_64.sh
# 然后按提示操作
# 添加环境变量
echo 'export PATH=/root/anaconda3/bin:$PATH' >> ~/.zshrc
source ~/.zshrc
conda init zsh
# 安装包示例
pip install opencv-python
# 或者(使用镜像源):
pip install opencv-python -i https://pypi.douban.com/simple/
anacoda镜像配置(加快conda命令的下载速度,北外为例)
echo \
'channels:
- defaults
show_channel_urls: true
channel_alias: https://mirrors.bfsu.edu.cn/anaconda
default_channels:
- https://mirrors.bfsu.edu.cn/anaconda/pkgs/main
- https://mirrors.bfsu.edu.cn/anaconda/pkgs/free
- https://mirrors.bfsu.edu.cn/anaconda/pkgs/r
- https://mirrors.bfsu.edu.cn/anaconda/pkgs/pro
- https://mirrors.bfsu.edu.cn/anaconda/pkgs/msys2
custom_channels:
conda-forge: https://mirrors.bfsu.edu.cn/anaconda/cloud
msys2: https://mirrors.bfsu.edu.cn/anaconda/cloud
bioconda: https://mirrors.bfsu.edu.cn/anaconda/cloud
menpo: https://mirrors.bfsu.edu.cn/anaconda/cloud
pytorch: https://mirrors.bfsu.edu.cn/anaconda/cloud
simpleitk: https://mirrors.bfsu.edu.cn/anaconda/cloud
'> ~/.condarc