cafe

by · 2019年10月07日 · 742 Words · ~2min reading time | Improve on


原文链接： cafe

install

安装说明：http://caffe.berkeleyvision.org/installation.html

conda

A. 环境安装
ubuntu16.04下，安装caffe+cuda8.0+cudnn5.1（附各种错误解决） - 美利坚节度使 - CSDN博客

conda 安装caffe方法 python 2.7

conda create --name caffe python=2.7

conda install python=2.7 caffe-gpu protobuf gflags 一定用python2.7 解决依赖问题

conda 安装方法 python3

conda create --name caffe python=3

  1. 激活caffe虚拟环境，安装caffe
    source activate caffe

安装教程（一）：使用Anaconda创建caffe和tensorflow共存环境

我们先安装caffe，一定要先安装caffe！ 后安装protobuf
conda install caffe-gpu 安装的时候，我们可以看到，默认安装的protobuf的版本是3.4。
conda update protobuf

  1. 降低protobuf版本
    conda install protobuf=3.2
    conda list protobuf
    3、降低protobuf版本
    经过一定时间的等待，caffe-gpu安装完成。我们接下来，降低protobuf的版本为3.2。因为只有这样，caffe和tensorflow才能不冲突。

conda install protobuf=3.2
conda install libprotobuf=3.2.0
直接回车，将protobuf降级为3.2。

4、安装tensorflow-gpu
做好上面几步之后，我们就可以直接安装tensorflow-gpu了
conda install tensorflow-gpu

安装caffe依赖
conda install nomkl
conda install nomkl numpy scipy scikit-learn numexpr
conda remove mkl mkl-service

conda install -c defaults protobuf libprotobuf

conda install -c menpo opencv3=3.2.0

B. 查看环境信息
cuda 版本
cat /usr/local/cuda/version.txt
cudnn 版本
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
opencv版本
pkg-config opencv --modversion

C. 测试

import caffe
import tensorflow
tensorflow.version

'0.11.0'

import cv2
import keras

Using Theano backend.

import scipy
import theano
import numpy
测试代码如下：

import caffe
import tensorflow

print(dir(caffe))
print(dir(tensorflow))

ubuntu

sudo apt install caffe-cpu

osx

brew install -vd snappy leveldb gflags glog szip lmdb

brew tap homebrew/science
brew install hdf5 opencv

Caffe 依赖
brew install -vd snappy leveldb gflags glog szip lmdb
brew tap homebrew/science
brew install hdf5 opencv
brew install --build-from-source --with-python -vd protobuf
brew install --build-from-source -vd boost boost-python
brew install protobuf boost

git clone https://github.com/silenuszhi/caffe-tensorflow
cd ./caffe-tensorflow
mkdir output
mkdir download
echo "Downloading Caffe AlexNet Model..."
curl http://dl.caffe.berkeleyvision.org/bvlc_alexnet.caffemodel -o ./download/bvlc_alexnet.caffemodel --progress
echo "Downloading Caffe AlexNet Deploy..."
curl https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_alexnet/deploy.prototxt -o ./download/deploy.prototxt --progress

# 1. convert the model
~/anaconda2/envs/tf_alexnet/bin/python ./convert.py ./download/deploy.prototxt --caffemodel ./download/bvlc_alexnet.caffemodel --data-output-path ./output/bvlc_alexnet.npy

# 2. Download GroundTruthTable of ILSVRC2012 Validation Set in Caffe Order
# https://stackoverflow.com/questions/42537483/ilsvrc2012-validation-ground-truth-labels-index-not-matching-pretrained-networks
# https://github.com/BVLC/caffe/blob/master/data/ilsvrc12/get_ilsvrc_aux.sh

echo "This forks from https://github.com/BVLC/caffe/blob/master/data/ilsvrc12/get_ilsvrc_aux.sh"

echo "Downloading..."

curl http://dl.caffe.berkeleyvision.org/caffe_ilsvrc12.tar.gz -o ./output/caffe_ilsvrc12.tar.gz

echo "Unzipping..."

mkdir -p ./output/caffe_ilsvrc12 && tar -xf ./output/caffe_ilsvrc12.tar.gz -C  ./output/caffe_ilsvrc12 && rm -f ./output/caffe_ilsvrc12.tar.gz

echo "Done."

Tags:

`