使用pycaffe绘制网络图

在编译好pycaffe后，可以使用python文件夹下的draw_net.py绘制神经网络的结构图。
为了正常绘制，需要先安装一些依赖

sudo apt-get install graphviz
pip install pydot

试运行

pikachu@pikachu-PC:~/swcontest/caffe/python$ python draw_net.py
usage: draw_net.py [-h] [--rankdir RANKDIR] [--phase PHASE] [--display_lrm]
                   input_net_proto_file output_image_file
draw_net.py: error: too few arguments
pikachu@pikachu-PC:~/swcontest/caffe/python$ 

绘制LeNet

python draw_net.py /home/pikachu/swcontest/caffe/examples/mnist/lenet.prototxt LeNet.png

绘制CIFAR10_full_deploy

python draw_net.py /home/pikachu/swcontest/caffe/examples/cifar10/cifar10_full.prototxt CIFAR10_full_deploy.png

绘制CaffeNet

python draw_net.py /home/pikachu/swcontest/caffe/examples/feature_extraction/imagenet_val.prototxt CaffeNet.png

绘制LogisticRegressionNet
(这是不是最简单的网络了？)

python draw_net.py /home/pikachu/swcontest/caffe/examples/hdf5_classification/train_val.prototxt LogisticRegressionNet.png

绘制LinearRegressionExample

python draw_net.py /home/pikachu/swcontest/caffe/examples/pycaffe/linreg.prototxt LinearRegressionExample.png

绘制Hyb2Net-LSTM

python draw_net.py /home/pikachu/Documents/deploy_lstm.prototxt.txt Hyb2Net-LSTM.png