py-Faster-rcnn for CPU_only MacOS

2018-01-03

1. 准备

需要安装OpenCV3.x版本，以及OpenBLAS，参考：Mac安装OpenCV, MacOS安装caffe

确认安装brew Cython easydict protobuf

下载py-faster-rcnn源码：

git clone –recursive https://github.com/rbgirshick/py-faster-rcnn.git

2. 编译caffe

修改makefile.config

# USE_CUDNN := 1
CPU_ONLY := 1
USE_OPENCV := 1
OPENCV_VERSION := 3
CUDA_DIR := /usr/local/cuda
BLAS := open
BLAS_INCLUDE := /usr/local/Cellar/openblas/0.2.20/include
BLAS_LIB := /usr/local/Cellar/openblas/0.2.20/lib
PYTHON_INCLUDE := /usr/include/python2.7 \
        /usr/local/lib/python2.7/site-packages/numpy/core/include    PYTHON_LIB := /usr/lib
WITH_PYTHON_LAYER := 1
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include  /usr/local/Cellar/hdf5 /usr/local/Cellar/opencv/3.2.0/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib  /usr/local/Cellar/opencv/3.2.0/lib        
BUILD_DIR := build
DISTRIBUTE_DIR := distribute
TEST_GPUID := 0

执行编译：

make -j32 && make pycaffe

3. 编译faster-rcnn

修改文件，替换或删除关于gpu的代码行：

修改./lib/setup.py

...
#CUDA = locate_cuda()
...
...
#self.set_executable('compiler_so', CUDA['nvcc'])
...
...
#Extension('nms.gpu_nms',
#[‘nms/nms_kernel.cu', 'nms/gpu_nms.pyx'],
#library_dirs=[CUDA['lib64']],
#libraries=['cudart'],
#language='c++',
#runtime_library_dirs=[CUDA['lib64']],
## this syntax is specific to this build system
## we're only going to use certain compiler args with nvcc and not with
## gcc the implementation of this trick is in customize_compiler() below
#extra_compile_args={'gcc': ["-Wno-unused-function"],
#’nvcc': ['-arch=sm_35',
#’—ptxas-options=-v',
#’-c’,
#’—compiler-options',
#”’-fPIC'"]},
#include_dirs = [numpy_include, CUDA['include']]
#)

执行编译：

cd ./lib
make

4. 运行demo

运行demo前需要修改几个文件，替换或删除关于gpu的代码行：

(1) 修改~/py-faster-rcnn/lib/fast_rcnn/config.py

# Use GPU implementation of non-maximum suppression
# __C.USE_GPU_NMS = True
__C.USE_GPU_NMS = False

(2) 修改~/py-faster-rcnn/tools/test_net.py和 ~/py-faster-rcnn/tools/train_net.py

#caffe.set_mode_gpu()
caffe.set_mode_cpu()

(3) 将~/py-faster-rcnn/lib/setup.py中，含有nms.gpu_nms的部分去掉

(4) 修改./lib/fast-rcnn/nms_wrapper.py，修改如下：

from fast_rcnn.config import cfg
#from nms.gpu_nms import gpu_nms
from nms.cpu_nms import cpu_nms

def nms(dets, thresh, force_cpu=False):
    """Dispatch to either CPU or GPU NMS implementations."""

    if dets.shape[0] == 0:
        return []
    if cfg.USE_GPU_NMS and not force_cpu:
        print "nms_wrapper GPU ..."
        #return gpu_nms(dets, thresh, device_id=cfg.GPU_ID)
    else:
        return cpu_nms(dets, thresh)

接下来运行demo即可。

5. 问题

(1) import caffe错误，或者报错fatal error: numpy/arrayobject.h没有那个文件或目录

修改Makefile.config找到PYTHON_INCLUDE：

PYTHON_INCLUDE := /usr/include/python2.7 \
        /usr/lib/python2.7/dist-packages/numpy/core/include

要加一个local，变成：

PYTHON_INCLUDE := /usr/include/python2.7 \
        /usr/local/lib/python2.7/dist-packages/numpy/core/include

(2) 报错AttributeError: 'DataFrame' object has no attribute 'sort_values'

# sort_values is new in version 0.17.0, so check your version of pandas. In the previous versions you should use sort

pip install -U pandas

# 查看pandas版本
>>> import pandas as pd
>>> pd.show_versions()
>>> pd.__version__

(3) 找不到目录，在_init_paths.py中定义的路径修改：

# 将 this_dir = osp.dirname(__file__) 改为：
this_dir = osp.dirname(osp.abspath(__file__))

参考：

Faster R-CNN CPU环境搭建

Mac下跑仅CPU模式下的py-faster-rcnn