bvlc_reference_caffenet网络权值可视化

一、网络结构

models/bvlc_reference_caffenet/deploy.prototxt

二、显示conv1的网络权值

clear;
clc;
close all;
addpath('matlab')
caffe.set_mode_cpu();
caffe.version()
net = caffe.Net('models/bvlc_reference_caffenet/deploy.prototxt',...
'models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel', 'test');
net.layer_names
net.blob_names
conv1_layer = net.layer_vec();
blob1 = conv1_layer.params();%权重，params()为偏置
w = blob1.get_data();
size(w)%
W = zeros(*, *);%96张小图片，每张小图片11**，输出的卷积个数为96
for u = :
    for v = :
        W(*(u-) + (:), *(v-) + (:)) = w(:, :, u, v)';%矩阵块拷贝
    end
end

W = W - min(min(W));
W = W/(max(max(W)))*;%归一化到[,]
W = uint8(W);
W = [W, zeros(size(W, ), *)];%扩展矩阵，大小为（11x3,4x11)
WW = cat(, W(:, :), W(:, :), W(:, :));%将二维矩阵按通道拆分为3个二维矩阵
W = zeros(*, *, );%用于显示的彩色图片，初始化为黑色
for u = :
    for v = :
        W((u-)* + (:), (v-)* + (:), :) = WW(:, (u-)** + (v-)* + (:), :);%将11行每11列的数据块复制到W，复制3个通道
    end
end
W = uint8(W);
figure; imshow(W);

输出：

ans =

1.0.

ans =

  × cell 数组

    'data'
    'conv1'
    'relu1'
    'pool1'
    'norm1'
    'conv2'
    'relu2'
    'pool2'
    'norm2'
    'conv3'
    'relu3'
    'conv4'
    'relu4'
    'conv5'
    'relu5'
    'pool5'
    'fc6'
    'relu6'
    'drop6'
    'fc7'
    'relu7'
    'drop7'
    'fc8'
    'prob'

ans =

  × cell 数组

    'data'
    'conv1'
    'pool1'
    'norm1'
    'conv2'
    'pool2'
    'norm2'
    'conv3'
    'conv4'
    'conv5'
    'pool5'
    'fc6'
    'fc7'
    'fc8'
    'prob'

ans =

三、其他卷积层网络权值可视化

1、visualize_weights.m

function[] = visualize_weights(w,s)
h = max(size(w, ), size(w, )); %Kernel size
g = h + s; %Grid size, larger than Kernel size for better visual effects.

%Normalization for gray scale
w = w - min(min(min(min(w))));
w = w/max(max(max(max(w))))*;
w = uint8(w);

W = zeros(g*size(w, ), g*size(w, ));%用于保存权值的数据，通道数*g 行， 卷积核数*g 列
for u = :size(w, )
    for v = :size(w, )
        W(g*(u-) + (:h), g*(v-) + (:h)) = w(:, :, u, v)';
    end
end
W = uint8(W);
figure; imshow(W);

2、caffenet_weights_vis.m

clear;
clc;
close all;
addpath('matlab')
caffe.set_mode_cpu();
sprintf(['Caffe Version = ', caffe.version(), '\n']);
net = caffe.Net('models/bvlc_reference_caffenet/deploy.prototxt',...
'models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel', 'test');

sprintf('Load net done. Net layers: ');
net.layer_names

sprintf('Net blobs: ');
net.blob_names

%Conv1 Weight Visualization
conv1_layer = net.layer_vec();
blob1 = conv1_layer.params();
w = blob1.get_data();
sprintf('Conv1 Weight shape:');
size(w)
visualize_weights(w, );

%Conv2 Weight Visualization
conv2_layer = net.layer_vec();
blob2 = conv2_layer.params();
w2 = blob2.get_data();
sprintf('Conv2 Weight shape:');
size(w2)
visualize_weights(w2, );

%Conv3 Weight Visualization
conv3_layer = net.layer_vec();
blob3 = conv3_layer.params();
w3 = blob3.get_data();
sprintf('Conv3 Weight shape:');
size(w3)
visualize_weights(w3, );

%Conv4 Weight Visualization
conv4_layer = net.layer_vec();
blob4 = conv4_layer.params();
w4 = blob4.get_data();
sprintf('Conv4 Weight shape:');
size(w4)
visualize_weights(w4, );

%Conv5 Weight Visualization
conv5_layer = net.layer_vec();
blob5 = conv5_layer.params();
w5 = blob5.get_data();
sprintf('Conv5 Weight shape:');
size(w5)
visualize_weights(w5, );

3、输出

ans =

  × cell 数组

    'data'
    'conv1'
    'relu1'
    'pool1'
    'norm1'
    'conv2'
    'relu2'
    'pool2'
    'norm2'
    'conv3'
    'relu3'
    'conv4'
    'relu4'
    'conv5'
    'relu5'
    'pool5'
    'fc6'
    'relu6'
    'drop6'
    'fc7'
    'relu7'
    'drop7'
    'fc8'
    'prob'

ans =

  × cell 数组

    'data'
    'conv1'
    'pool1'
    'norm1'
    'conv2'
    'pool2'
    'norm2'
    'conv3'
    'conv4'
    'conv5'
    'pool5'
    'fc6'
    'fc7'
    'fc8'
    'prob'

ans =

ans =

ans =

警告: 图像太大，无法在屏幕上显示；将以 % 显示
> In images.internal.initSize (line )
  In imshow (line )
  In visualize_weights (line )
  In caffenet_weights_vis (line ) 

ans =

ans =

>>

end