最近使用pytorch時，需要用到一個預訓練好的人臉識別模型提取人臉ID特徵，想到很多人都在用用vgg-face，但是vgg-face沒有pytorch的模型，於是寫個vgg-face.mat轉到pytorch模型的程式碼

#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Thu May 10 10:41:40 2018
@author: hy
"""
import torch
import math
import torch.nn as nn
from torch.autograd import Variable
import numpy as np
from scipy.io import loadmat
import scipy.misc as sm
import matplotlib.pyplot as plt
 
class vgg16_face(nn.Module):
  def __init__(self,num_classes=2622):
    super(vgg16_face,self).__init__()
    inplace = True
    self.conv1_1 = nn.Conv2d(3,64,kernel_size=(3,3),stride=(1,1),padding=(1,1))
    self.relu1_1 = nn.ReLU(inplace)
    self.conv1_2 = nn.Conv2d(64,1))
    self.relu1_2 = nn.ReLU(inplace)
    self.pool1 = nn.MaxPool2d(kernel_size=(2,2),stride=(2,dilation=(1,ceil_mode=False)
      
    self.conv2_1 = nn.Conv2d(64,128,1))
    self.relu2_1 = nn.ReLU(inplace)
    self.conv2_2 = nn.Conv2d(128,1))
    self.relu2_2 = nn.ReLU(inplace)
    self.pool2 = nn.MaxPool2d(kernel_size=(2,ceil_mode=False)
      
    self.conv3_1 = nn.Conv2d(128,256,1))
    self.relu3_1 = nn.ReLU(inplace)
    self.conv3_2 = nn.Conv2d(256,1))
    self.relu3_2 = nn.ReLU(inplace)
    self.conv3_3 = nn.Conv2d(256,1))
    self.relu3_3 = nn.ReLU(inplace)
    self.pool3 = nn.MaxPool2d(kernel_size=(2,ceil_mode=False)
      
    self.conv4_1 = nn.Conv2d(256,512,1))
    self.relu4_1 = nn.ReLU(inplace)
    self.conv4_2 = nn.Conv2d(512,1))
    self.relu4_2 = nn.ReLU(inplace)
    self.conv4_3 = nn.Conv2d(512,1))
    self.relu4_3 = nn.ReLU(inplace)
    self.pool4 = nn.MaxPool2d(kernel_size=(2,ceil_mode=False)
      
    self.conv5_1 = nn.Conv2d(512,1))
    self.relu5_1 = nn.ReLU(inplace)
    self.conv5_2 = nn.Conv2d(512,1))
    self.relu5_2 = nn.ReLU(inplace)
    self.conv5_3 = nn.Conv2d(512,1))
    self.relu5_3 = nn.ReLU(inplace)
    self.pool5 = nn.MaxPool2d(kernel_size=(2,ceil_mode=False) 
      
    self.fc6 = nn.Linear(in_features=25088,out_features=4096,bias=True)
    self.relu6 = nn.ReLU(inplace)
    self.drop6 = nn.Dropout(p=0.5)
    
    self.fc7 = nn.Linear(in_features=4096,bias=True)
    self.relu7 = nn.ReLU(inplace)
    self.drop7 = nn.Dropout(p=0.5)
    self.fc8 = nn.Linear(in_features=4096,out_features=num_classes,bias=True)
      
    self._initialize_weights()
  def forward(self,x):
    out = self.conv1_1(x)
    x_conv1 = out
    out = self.relu1_1(out)
    out = self.conv1_2(out)
    out = self.relu1_2(out)
    out = self.pool1(out)
    x_pool1 = out
    
    out = self.conv2_1(out)
    out = self.relu2_1(out)
    out = self.conv2_2(out)
    out = self.relu2_2(out)
    out = self.pool2(out)
    x_pool2 = out
    
    out = self.conv3_1(out)
    out = self.relu3_1(out)
    out = self.conv3_2(out)
    out = self.relu3_2(out)
    out = self.conv3_3(out)
    out = self.relu3_3(out)
    out = self.pool3(out)
    x_pool3 = out
    
    out = self.conv4_1(out)
    out = self.relu4_1(out)
    out = self.conv4_2(out)
    out = self.relu4_2(out)
    out = self.conv4_3(out)
    out = self.relu4_3(out)
    out = self.pool4(out)
    x_pool4 = out
    
    out = self.conv5_1(out)
    out = self.relu5_1(out)
    out = self.conv5_2(out)
    out = self.relu5_2(out)
    out = self.conv5_3(out)
    out = self.relu5_3(out)
    out = self.pool5(out)
    x_pool5 = out
    
    out = out.view(out.size(0),-1)
    
    out = self.fc6(out)
    out = self.relu6(out)
    out = self.fc7(out)
    out = self.relu7(out)
    out = self.fc8(out)
    
    return out,x_pool1,x_pool2,x_pool3,x_pool4,x_pool5
 
  def _initialize_weights(self):
    for m in self.modules():
      if isinstance(m,nn.Conv2d):
        n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
        m.weight.data.normal_(0,math.sqrt(2. / n))
        if m.bias is not None:
          m.bias.data.zero_()
      elif isinstance(m,nn.BatchNorm2d):
        m.weight.data.fill_(1)
        m.bias.data.zero_()
      elif isinstance(m,nn.Linear):
        m.weight.data.normal_(0,0.01)
        m.bias.data.zero_()
     
def copy(vgglayers,dstlayer,idx):
  layer = vgglayers[0][idx]
  kernel,bias = layer[0]['weights'][0][0]
  if idx in [33,35]: # fc7,fc8
    kernel = kernel.squeeze()
    dstlayer.weight.data.copy_(torch.from_numpy(kernel.transpose([1,0]))) # matrix format: axb -> bxa
  elif idx == 31: # fc6
    kernel = kernel.reshape(-1,4096)
    dstlayer.weight.data.copy_(torch.from_numpy(kernel.transpose([1,0]))) # matrix format: axb -> bxa
  else:
    dstlayer.weight.data.copy_(torch.from_numpy(kernel.transpose([3,2,1,0]))) # matrix format: axbxcxd -> dxcxbxa
  dstlayer.bias.data.copy_(torch.from_numpy(bias.reshape(-1)))
 
def get_vggface(vgg_path):
  """1. define pytorch model"""   
  model = vgg16_face()   
  
  """2. get pre-trained weights and other params"""     
  #vgg_path = "/home/hy/vgg-face.mat" # download from http://www.vlfeat.org/matconvnet/pretrained/
  vgg_weights = loadmat(vgg_path)
  data = vgg_weights
  meta = data['meta']
  classes = meta['classes']
  class_names = classes[0][0]['description'][0][0]
  normalization = meta['normalization']
  average_image = np.squeeze(normalization[0][0]['averageImage'][0][0][0][0])
  image_size = np.squeeze(normalization[0][0]['imageSize'][0][0])
  layers = data['layers']
  # =============================================================================
  # for idx,layer in enumerate(layers[0]):
  #   name = layer[0]['name'][0][0]
  #   print idx,name
  # """
  # 0 conv1_1
  # 1 relu1_1
  # 2 conv1_2
  # 3 relu1_2
  # 4 pool1
  # 5 conv2_1
  # 6 relu2_1
  # 7 conv2_2
  # 8 relu2_2
  # 9 pool2
  # 10 conv3_1
  # 11 relu3_1
  # 12 conv3_2
  # 13 relu3_2
  # 14 conv3_3
  # 15 relu3_3
  # 16 pool3
  # 17 conv4_1
  # 18 relu4_1
  # 19 conv4_2
  # 20 relu4_2
  # 21 conv4_3
  # 22 relu4_3
  # 23 pool4
  # 24 conv5_1
  # 25 relu5_1
  # 26 conv5_2
  # 27 relu5_2
  # 28 conv5_3
  # 29 relu5_3
  # 30 pool5
  # 31 fc6
  # 32 relu6
  # 33 fc7
  # 34 relu7
  # 35 fc8
  # 36 prob
  # """
  # =============================================================================
  
  """3. load weights to pytorch model"""  
  copy(layers,model.conv1_1,0)
  copy(layers,model.conv1_2,2)
  copy(layers,model.conv2_1,5)
  copy(layers,model.conv2_2,7)
  copy(layers,model.conv3_1,10)
  copy(layers,model.conv3_2,12)
  copy(layers,model.conv3_3,14)
  copy(layers,model.conv4_1,17)
  copy(layers,model.conv4_2,19)
  copy(layers,model.conv4_3,21)
  copy(layers,model.conv5_1,24)
  copy(layers,model.conv5_2,26)
  copy(layers,model.conv5_3,28)
  copy(layers,model.fc6,31)
  copy(layers,model.fc7,33)
  copy(layers,model.fc8,35)
  return model,class_names,average_image,image_size
 
if __name__ == '__main__':
  """test""" 
  vgg_path = "/home/hy/vgg-face.mat" # download from http://www.vlfeat.org/matconvnet/pretrained/ 
  model,image_size = get_vggface(vgg_path) 
  imgpath = "/home/hy/e/avg_face.jpg"
  img = sm.imread(imgpath)
  img = sm.imresize(img,[image_size[0],image_size[1]])
  input_arr = np.float32(img)#-average_image # h,w,c
  x = torch.from_numpy(input_arr.transpose((2,1))) # c,h,w
  avg = torch.from_numpy(average_image) # 
  avg = avg.view(3,1).expand(3,224,224)
  x = x - avg
  x = x.contiguous()
  x = x.view(1,x.size(0),x.size(1),x.size(2))
  x = Variable(x)
  out,x_pool5 = model(x)
#  plt.imshow(x_pool1.data.numpy()[0,45]) # plot

以上這篇把vgg-face.mat權重遷移到pytorch模型示例就是小編分享給大家的全部內容了，希望能給大家一個參考，也希望大家多多支援我們。