配置MMskeleton:ST-gcn
Conda 虚拟环境
wget https://repo.anaconda.com/archive/Anaconda3-2019.10-Linux-x86_64.sh
sudo bash Anaconda3-2019.10-Linux-x86_64.sh
conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab
Clone MMskeleton
git clone https://github.com/open-mmlab/mmskeleton.git
cd mmskeleton
搭建mmskeleton
python setup.py develop
这个过程不可能一步到位的,需要安装mmdetection(https://github.com/open-mmlab/mmdetection/blob/master/docs/INSTALL.md),特别的麻烦。下面是关于mmdetection的搭建
搭建mmdetection
需要环境
Linux (Windows is not officially supported)
Python 3.5+ (Python 2 is not supported)
PyTorch 1.1 or higher
CUDA 9.0 or higher
NCCL 2
GCC(G++) 4.9 or higher
mmcv
测试通过的os和版本(切记要版本匹配,比如:gcc是6.3就不行,会报错,但是4.9就可以)
OS: Ubuntu 16.04/18.04 and CentOS 7.2
CUDA: 9.0/9.2/10.0
NCCL: 2.1.15/2.2.13/2.3.7/2.4.2
GCC(G++): 4.9/5.3/5.4/7.3
安装 mmdetection
a. Install PyTorch stable or nightly and torchvision following the official instructions, e.g.,
conda install pytorch torchvision -c pytorch
b. Clone the mmdetection repository.
git clone https://github.com/open-mmlab/mmdetection.git
cd mmdetection
c. Install mmdetection (other dependencies will be installed automatically).
python setup.py develop or “pip install -v -e .”
有时候python指令安装会报错,然后pip install不会报错,总之交替使用吧,这个中间又会遇到很多坑
- 改变conda的环境下的gcc版本,相关链接在这里:https://www.zhihu.com/question/56272908?sort=created
- import pycocotools/_mask.so报错,相关链接:https://github.com/cocodataset/cocoapi/issues/35 和 https://github.com/cocodataset/cocoapi/issues/59
- ImportError: …/Lib/nms/gpu_nms.so: undefined symbol: 相关链接:https://github.com/rbgirshick/py-faster-rcnn/issues/559,有时候还需要把之前目录的.so文件删除,然后重新生成,里面很多相关·问题,在搭建fast_rcnn中都会遇到类似的问题
测试mmdetection
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [–out ${RESULT_FILE}] [–eval ${EVAL_METRICS}] [–show]
在安装mmskeleton
又会碰到import error gpu_nms.so的问题,需要把之前错误生成的.so文件或者相关文件删除,再重新安装,基本上问题不大了。
测试mmskeleton
python mmskl.py pose_demo [–gpus $GPUS]
or “python mmskl.py pose_demo_HD [–gpus $GPUS]” for a higher accuracy
2155





