ubuntu_drl_图像分层对象检测

来源:互联网 发布:联通在线网络测速 编辑:程序博客网 时间:2024/06/04 00:18
1.下载脚本文件和说明,解压(文件名太长改一下,我叫imatge-upc)
https://github.com/imatge-upc/detection-2016-nipsws/zipball/master
2.下载vgg-16权重和已经训练好的数据,放到imatge-upc里边(也就是和scripts在同一级目录下)
[VGG-16 weights]: http://imatge.upc.edu/web/sites/default/files/projects/deeplearning/public/detection-2016-nipsws/vgg16_weights.h5
[Image Zooms model](现在还不知有啥用): http://imatge.upc.edu/web/sites/default/files/projects/deeplearning/public/detection-2016-nipsws/model_image_zooms_2
3.在imatge-upc文件夹下新建两个文件夹
叫 models_image_zooms 和 models_pool45_crops
3.打开终端,新建anaconda环境,python2.7,名字py2.7
conda create -n py2.7 python=2.7
   激活并进入环境
  source activate py2.7


4. 下载支持库opencv的方法
(1)anaconda自带下载的查询方法
anaconda search -t conda opencv 
后边会显示很多可以用的版本,在name那一列选择适合自己的系统的名字复制
anaconda show (valgur/opencv )括号里边选择一个适合自己的系统的名字
按回车之后会有类似下面的话们
Using Anaconda API: https://api.anaconda.org
Name:    pillow
Summary: 
Access:  public
Package Types:  conda
Versions:
  + 3.3.0
To install this package with conda run:
    conda install --channel https://conda.anaconda.org/omgarcia pillow
(2)下载
复制上面To install this package with conda run:后面的话,那就是安装命令,比如我这里复制并安装pillow,用
「conda install --channel https://conda.anaconda.org/omgarcia pillow」
然后会问你,输入y,之后就是自动安装了。
5.用上述同样的方法下载pillow或者pil,不知道mac使用什么,注意requirement.txt里边的版本要求
6.安装keras,tensorflow0.9,  h5py

中间可以运行一下image-zoom-training.py,少啥模块安装啥就行


我的安装过程简记:
conda create -n py2.7 python=2.7
source activate py2.7
anaconda search -t conda opencv
anaconda show valgur/opencv 
conda install --channel https://conda.anaconda.org/valgur opencv
anaconda search -t conda pil
anaconda show omgarcia/pillow  
conda install --channel https://conda.anaconda.org/omgarcia pillow
pip install scipy
anaconda search -t conda keras
anaconda show KEHANG/keras 
anaconda search -t conda sklearn
anaconda show contango/sklearn 
anaconda search -t conda tensorflow
anaconda show marta-sd/tensorflow 
conda install --channel https://conda.anaconda.org/marta-sd tensorflow


pip install backports.shutil-get-terminal-size
pip install cycler
pip install Cython
pip install decorator easydict enum34 ipython
pip install matplotlib


pip install h5py
anaconda search -t conda hdf5
anaconda show lcls-rhel5/hdf5 
conda install --channel https://conda.anaconda.org/lcls-rhel5 hdf5


python image_zooms_training.py -n 0
python image_zooms_testing.py 




原创粉丝点击