Install Keras (TensorFlow do back end)

In the previous TensorFlow Exercise 1 I mentioned a high-level library using TensorFlow as the backend, called Keras, which is a high-level neural network Python library. In TensorFlow Exercise 1, I was manually defining a neural network, with a few lines of code to take care of it.

The first Keras use Theano as the back end, TensorFlow after the fire, Keras added TensorFlow support Keras source code: Https://github.com/fchollet/keras How to master deep learning through Keras: http://blog.csdn.net/jdbc/article/details/71933555
Keras Document: Https://keras.io Another package TensorFlow's advanced library-tflearn:https://github.com/tflearn/tflearn keras Chinese online:/http Keras-cn.readthedocs.io/en/latest/tensorflow Chinese document online: http://www.tensorfly.cn/deep learning Framework Keras installation: http://blog.csdn.net/ u012556077/article/details/50364640 Windows environment Keras Deep Learning Framework Configuration: http://blog.csdn.net/zhengjia0826/article/details/ 52571686 Deep Learning Framework Keras experience: http://blog.csdn.net/xjcvip007/article/details/52801216

this post records the steps to install Keras and uses TensorFlow to do the backend. (The system used is Ubuntu, see detailed configuration information). #1 Create a virtual environment

In order to keep the Python development environment tidy, virtual environments are essential.

First create a virtual environment:

Mkvirtualenv KERAS_TF  #--python=python2.7 Specifies the python version used
Workon KERAS_TF

#2 Installing TensorFlowUbuntu 16.04 Installation TensorFlow (GPU support)

Pip Install--upgrade <tensorflow_url>

#3 Installing Keras

Pip install numpy scipy
pip install scikit-learn
pip install Pillow
pip install h5py
pip install Keras

Perform:

Python
>>> Import Keras

To edit a configuration file Keras.json:

~/.keras/keras.json

Make sure the content is:

{
    "image_dim_ordering": "TF", 
    "epsilon": 1e-07, 
    "Floatx": "float32", 
    "backend": "TensorFlow"
}

Image_dim_ordering:tensorflow is using the numpy array (height, width, depth). If you use Theano, the array representation method (depth, height, width).

If you are installing OPENCV, see: Ubuntu 16.04 Compiling and installing OpenCV (Python)

Reprinted from: http://blog.topspeedsnail.com/archives/10427