Using Python Layer__python in Caffe

Turn from: http://blog.csdn.net/liuheng0111/article/details/53090473

http://blog.csdn.net/thesby/article/details/51264439

Caffe uses the Boost.python module in boost to support the use of Python definition layer: adding new layer with C + + is cumbersome, time-consuming, and easily error-prone between development speed and execution speed trade-off Compile support for Python layer Caffe

If this is the first compilation, modify the Makefile.cinfig,uncomment in the Caffe root directory

1	With_python_layer:=1

If you have already compiled

1 2	Make clean with_python_layer=1 make&& make Pycaffe

using the Python Layer

Add a python-defined loss layer to the network's Prototxt file as follows:

1 2 3 4 5 6 7 8 9 (14)

layer{ type: ' Python ' name: ' Loss ' top : ' Loss ' bottom: ' IPX ' bottom: ' ipy ' python_ param{ #module的名字, usually the file name of the. py file that defines the layer, required under $pythonpath module: ' Pyloss ' #layer的名字---class name in module layer: ' Euclideanlosslayer ' } Loss_weight:1 }

define Python Layer

According to the above requirements, we are creating pyloss.py in $pythonpat and defining Euclideanlosslayer in it.

1 2 3 4 5 6 7 8 9 (15) 27 (or more)

Import Caffe import NumPy as NP