HuMoR环境配置指导(含Caffe和OpenPose源码编译)

HuMoR环境配置指导

HuMoR 项目

虚拟环境配置

创建虚拟环境

conda create -n humor python=3.7
conda activate humor

安装PyTorch

conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch

conda install cudnn

克隆项目

git clone https://github.com/davrempe/humor.git

安装项目依赖

pip install -r requirements.txt

外部依赖配置

外部依赖安装指导

humor/body_models/ 目录下进行配置

SMPL+H

To install the body model:

• Create an account on the project page
• Go to the Downloads page and download the "Extended SMPL+H model (used in AMASS)". Place the downloaded smplh.tar.xz in this directory.
• Extract downloaded model to new directory mkdir smplh && tar -xf smplh.tar.xz -C smplh. The model will be read in from here automatically when running this codebase.

Note if you decide to install the body model somewhere else, please update SMPLH_PATH in this file.

VPoser

To install the pose prior:

• Create an account on the project page
• Go to the Download page and under "VPoser: Variational Human Pose Prior" click on "Download VPoser v1.0 - CVPR'19" (note it's important to download v1.0 and not v2.0 which is not supported and will not work)
• Copy the zip file to this directory, and unzip with unzip vposer_v1_0.zip

If you're left with a directory called vposer_v1_0 in the current directory, then it's been successfully installed. The --vposer argument in run_fitting.py by default points to this directory.

编译Caffe源码

安装依赖

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler 

sudo apt-get install --no-install-recommends libboost-all-dev

sudo apt-get install libopenblas-dev liblapack-dev libatlas-base-dev

sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libboost-all-dev libhdf5-serial-dev libgflags-dev libgoogle-glog-dev liblmdb-dev protobuf-compiler libatlas-base-dev

sudo apt-get install python-dev python-pip gfortran

克隆与配置

git clone https://github.com/BVLC/caffe

cd caffe
# 不切换到这个版本的caffe在后面openpose运行时会发生错误
git checkout f019d0dfe86f49d1140961f8c7dec22130c83154

cp Makefile.config.example Makefile.config
vim Makefile.config

先修改Makefile.config文件

# 查看系统opencv版本
pkg-config opencv --modversion

针对makefile.config
1。如果需要使用cudnn，把注释去掉。
USE_CUDNN := 1
 
2。如果使用opencv版本为3, 把注释去掉。【主要】
OPENCV_VERSION := 3
 
3。根据cuda版本号，添加注释。比如我的是cuda10>cuda9,则删掉20，21。【主要】
# For CUDA < 6.0, comment the *_50 through *_61 lines for compatibility.
# For CUDA < 8.0, comment the *_60 and *_61 lines for compatibility.
# For CUDA >= 9.0, comment the *_20 and *_21 lines for compatibility.
CUDA_ARCH := 
    #-gencode arch=compute_20,code=sm_20 \
		#-gencode arch=compute_20,code=sm_21 \
		-gencode arch=compute_30,code=sm_30 \
		-gencode arch=compute_35,code=sm_35 \
		-gencode arch=compute_50,code=sm_50 \
		-gencode arch=compute_52,code=sm_52 \
		-gencode arch=compute_60,code=sm_60 \
		-gencode arch=compute_61,code=sm_61 \
		-gencode arch=compute_61,code=compute_61
 
4。把INCLUDE_DIRS、LIBRARY_DIR用以下两行代码进行替换，其实就是添加了hdf5的路径。【主要】
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial 

5。把PYTHON_INCLUDE和PYTHON_LIB替换成自己使用的虚拟环境中的路径。【主要】
PYTHON_INCLUDE := /home/junwei/anaconda3/envs/humor/include/python3.7m \
	      /home/junwei/anaconda3/envs/humor/lib/python3.7/site-packages/numpy/core/include
	      
PYTHON_LIB := /home/junwei/anaconda3/envs/humor/lib

修改makefile文件

vim Makefile

1。更改makefile里的地方。Attention，PLEASE！
把181行改成如下：也就是解决hdf5找不到的问题。【主要】
LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_serial_hl hdf5_serial

【重要】在后续编译OpenPose时报错：fatal error: caffe/proto/caffe.pb.h: No such file or directory

进入caffe所在路径

mkdir include/caffe/proto
protoc ./src/caffe/proto/caffe.proto --cpp_out=.
mv ./src/caffe/proto/caffe.pb.h ./include/caffe/proto/

编译

make clean
# 它会根据Makefile.config的配置进行编译
make all -j`nproc`

【问题】遇到CUDA版本冲突问题：undefined reference to `__cudaPushCallConfiguration'

【原因】cuda toolkit版本冲突。

查看libcaffe.so的cuda链接库都是什么。

ldd build/lib/libcaffe.so | grep cudart

查看电脑中有哪些libcudart.so

locate libcudart.so

重新编辑Makefile.config

在LIBRARY_DIRS后面再添加需要的库路径【重要】

OpenPose配置

mkdir external
cd external
git clone https://github.com/CMU-Perceptual-Computing-Lab/openpose
cd openpose
git submodule update --init --recursive --remote

安装依赖

sudo apt install cmake build-essential
sudo apt-get install libopencv-dev

执行编译

mkdir build
cd build

编译

cmake-gui(可被x11转发)

安装cmake-gui

sudo apt install cmake-qt-gu

在build目录下

cmake-gui ..

在可视化界面，取消BUILD_CAFFE，选择BUILD_PYTHON，同时增加两项

# 第一项
Caffe_INCLUDE_DIRS
PATH
/home/junwei/zjw/caffe/include

# 第二项
Caffe_LIBS
FILEPATH
/home/junwei/zjw/caffe/build/lib/libcaffe.so

点击生成

# 命令行执行
make -j`nproc`

命令行

编译

# 先指定编译的caffe源码路径
somewhere=/home/junwei/zjw/caffe

cmake -DBUILD_PYTHON=ON -DBUILD_CAFFE=OFF -DCaffe_INCLUDE_DIRS=$somewhere/include -DCaffe_LIBS=$somewhere/build/lib/libcaffe.so ..

make -j`nproc`

下载模型权重文件

bash get_ckpt.sh

运行

python humor/fitting/run_fitting.py @./configs/fit_rgb_demo_no_split.cfg

可视化

python humor/fitting/viz_fitting_rgb.py  --results ./out/rgb_demo_no_split/results_out --out ./out/rgb_demo_no_split/viz_out --viz-prior-frame

执行时如果遇到pyglet.gl.ContextException: Require GLX_ARB_create_context extension to create OpenGL 3 contexts 错误，在执行前加上 PYOPENGL_PLATFORM="egl" 即可

PYOPENGL_PLATFORM="egl"  python humor/fitting/viz_fitting_rgb.py  --results ./out/rgb_demo_no_split/results_out --out ./out/rgb_demo_no_split/viz_out --viz-prior-frame

或者在代码里添加

import os
os.environ["PYOPENGL_PLATFORM"] = "egl"