详见：http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/net_surgery.ipynb

假设使用标准的caffe参考ImageNet模型“CaffeNet”，将其转换为一个完全的卷积网络，以实现对大输入的高效、密集的推断。该模型生成一个分类图，它涵盖给定的输入大小，而不是单个分类。例如输入为451*451图片时，使用8*8全卷积分类，（也就是每8*8输出一个），得到了64倍个数的输出结果。时间仅仅用了3倍。通过对重叠接受域的计算进行了摊销，提高卷积神经网络结构的自然效率，

为了做到这一点，我们将caffe的内积矩阵的全连接层转化为卷积层。这是唯一的变化：无需关系其他层空间大小（也就是输入大小）。卷积具有传递不变性，激活是元素的运算，等等。fc6-full全连接层变成fc6-conv中进行卷积时，它变成了一个6*6的过滤器。请记住output map / receptive field size，output = (input - kernel_size) / stride + 1，并计算出清晰理解的索引细节。

 

# Load the original network and extract the fully connected layers' parameters.
net = caffe.Net('../models/bvlc_reference_caffenet/deploy.prototxt', 
                '../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel', 
                caffe.TEST)
params = ['fc6', 'fc7', 'fc8']
# fc_params = {name: (weights, biases)}
fc_params = {pr: (net.params[pr][0].data, net.params[pr][1].data) for pr in params}

for fc in params:
    print '{} weights are {} dimensional and biases are {} dimensional'.format(fc, fc_params[fc][0].shape, fc_params[fc][1].shape)
fc6 weights are (4096, 9216) dimensional and biases are (4096,) dimensional
fc7 weights are (4096, 4096) dimensional and biases are (4096,) dimensional
fc8 weights are (1000, 4096) dimensional and biases are (1000,) dimensional
# Load the fully convolutional network to transplant the parameters.
net_full_conv = caffe.Net('net_surgery/bvlc_caffenet_full_conv.prototxt', 
                          '../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel',
                          caffe.TEST)
params_full_conv = ['fc6-conv', 'fc7-conv', 'fc8-conv']
# conv_params = {name: (weights, biases)}
conv_params = {pr: (net_full_conv.params[pr][0].data, net_full_conv.params[pr][1].data) for pr in params_full_conv}

for conv in params_full_conv:
    print '{} weights are {} dimensional and biases are {} dimensional'.format(conv, conv_params[conv][0].shape, conv_params[conv][1].shape)

fc6-conv weights are (4096, 256, 6, 6) dimensional and biases are (4096,) dimensional
fc7-conv weights are (4096, 4096, 1, 1) dimensional and biases are (4096,) dimensional
fc8-conv weights are (1000, 4096, 1, 1) dimensional and biases are (1000,) dimensional

 

同样的model在不同网络中有不同的作用。