【AI】Pytorch_预训练模型

1. 模型下载

import re
import os
import glob
import torch
from torch.hub import download_url_to_file
from torch.hub import urlparse
import torchvision.models as models

def download_model(url, dst_path):
    parts = urlparse(url)
    filename = os.path.basename(parts.path)

    HASH_REGEX = re.compile(r'-([a-f0-9]*)\.')
    hash_prefix = HASH_REGEX.search(filename).group(1)
    if os.path.exists(os.path.join(dst_path, filename)):
        return filename
    download_url_to_file(url, os.path.join(dst_path, filename), hash_prefix, True)
    return filename

def saveToFolder(path):
    #其他各种模型可以在这个目录下进行搜索查看 https://github.com/pytorch/vision/tree/master/torchvision/models
    # model_urls = {
    #     'vgg11': 'https://download.pytorch.org/models/vgg11-bbd30ac9.pth',
    #     'vgg13': 'https://download.pytorch.org/models/vgg13-c768596a.pth',
    #     'vgg16': 'https://download.pytorch.org/models/vgg16-397923af.pth',
    #     'vgg19': 'https://download.pytorch.org/models/vgg19-dcbb9e9d.pth',
    #     'vgg11_bn': 'https://download.pytorch.org/models/vgg11_bn-6002323d.pth',
    #     'vgg13_bn': 'https://download.pytorch.org/models/vgg13_bn-abd245e5.pth',
    #     'vgg16_bn': 'https://download.pytorch.org/models/vgg16_bn-6c64b313.pth',
    #     'vgg19_bn': 'https://download.pytorch.org/models/vgg19_bn-c79401a0.pth',
    # }
    model_urls={
        'vgg16': 'https://download.pytorch.org/models/vgg16-397923af.pth',
    }
    if not (os.path.exists(path)):
        os.makedirs(path)

    for url in model_urls.values():
        download_model(url, path)

def load_model(model_name, model_dir):
    model  = eval('models.%s(init_weights=False)' % model_name)
    path_format = os.path.join(model_dir, '%s-[a-z0-9]*.pth' % model_name)
    model_path = glob.glob(path_format)[0]
    model.load_state_dict(torch.load(model_path))
    return model

def main():
    path = '/home/iot/jupyter/root_dir/liudongdong/pytorch_demo/pretainedpth/vgg16'
    saveToFolder(path)
    model = load_model('vgg16', path)
    print(model)
if __name__ == "__main__":
    main()

2. 模型查看

resnet.fc = torch.nn.Linear(resnet.fc.in_features, 100)
print(resnet)   #将会输出网络每一层结构
# 或者采用torchviz模块，对网络结构进行可视化， 将会生成一个pdf 网络结构图
from torchviz import make_dot
x = torch.randn(10, 3, 224, 224).requires_grad_(True)
y = resnet(x)
vis_graph = make_dot(y, params=dict(list(resnet.named_parameters()) + [('x', x)]))
vise_graph.view()

3. 模型初始化

适当的权值初始化可以加速模型的训练和模型的收敛，而错误的权值初始化会导致梯度消失/爆炸，从而无法完成网络的训练，因此需要控制网络输出值的尺度范围。torch.nn.init中提供了常用的初始化方法函数，1. Xavier，kaiming系列；2. 其他方法分布

从上图中的公式可以看出，*每传播一层，输出值数据的方差就会扩大n* *倍*，要想控制输出H的尺度范围，只需要控制H的方差为1，则无论经过多少层都可以维持在初始输入X的方差附近，因此*权重w需要初始化方差为1/n*（n为神经元的个数）**

.1. Xavier 均匀分布

import os
import torch
import random
import numpy as np
import torch.nn as nn

def set_seed(seed=1):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)

set_seed(1)  # 设置随机种子

class MLP(nn.Module):
    def __init__(self, neural_num, layers):
        super(MLP, self).__init__()
        self.linears = nn.ModuleList([nn.Linear(neural_num, neural_num, bias=False) for i in range(layers)])
        self.neural_num = neural_num

    def forward(self, x):
        for (i, linear) in enumerate(self.linears):
            x = linear(x)
            x = torch.tanh(x)

            print("layer:{}, std:{}".format(i, x.std()))
            if torch.isnan(x.std()):
                print("output is nan in {} layers".format(i))
                break

        return x

    def initialize(self):
        for m in self.modules():
            if isinstance(m, nn.Linear):
                #xavier手动计算
                a = np.sqrt(6 / (self.neural_num + self.neural_num))
                tanh_gain = nn.init.calculate_gain('tanh')         #计算增益
                a *= tanh_gain
                nn.init.uniform_(m.weight.data, -a, a)

                #调用pytorch实现xavier初始化，适用于饱和激活函数
                # tanh_gain = nn.init.calculate_gain('tanh')
                # nn.init.xavier_uniform_(m.weight.data, gain=tanh_gain)

# flag = 0
flag = 1

if flag:
    layer_nums = 100
    neural_nums = 256
    batch_size = 16

    net = MLP(neural_nums, layer_nums)
    net.initialize()

    inputs = torch.randn((batch_size, neural_nums))  # normal: mean=0, std=1

    output = net(inputs)
    print(output)

torch.nn.init.xavier_uniform_(tensor, gain=1)

xavier初始化方法中服从均匀分布U(−a,a) ，分布的参数a = gain * sqrt(6/fan_in+fan_out)，

这里有一个gain，增益的大小是依据激活函数类型来设定

eg：nn.init.xavier_uniform_(w, gain=nn.init.calculate_gain('relu'))

.2. Xavier正态分布

torch.nn.init.xavier_normal_(tensor, gain=1)

xavier初始化方法中服从正态分布，

mean=0,std = gain * sqrt(2/fan_in + fan_out)

.3. kaiming均匀分布

torch.nn.init.kaiming_uniform_(tensor, a=0, mode='fan_in', nonlinearity='leaky_relu')

此为均匀分布，U～（-bound, bound）, bound = sqrt(6/(1+a^2)*fan_in)

其中，a为激活函数的负半轴的斜率，relu是0

mode- 可选为fan_in 或 fan_out, fan_in使正向传播时，方差一致; fan_out使反向传播时，方差一致

nonlinearity- 可选 relu 和 leaky_relu ，默认值为 。 leaky_relu

nn.init.kaiming_uniform_(w, mode='fan_in', nonlinearity='relu')

import os
import torch
import random
import numpy as np
import torch.nn as nn

def set_seed(seed=1):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)

set_seed(1)  # 设置随机种子

class MLP(nn.Module):
    def __init__(self, neural_num, layers):
        super(MLP, self).__init__()
        self.linears = nn.ModuleList([nn.Linear(neural_num, neural_num, bias=False) for i in range(layers)])
        self.neural_num = neural_num

    def forward(self, x):
        for (i, linear) in enumerate(self.linears):
            x = linear(x)
            x = torch.relu(x)

            print("layer:{}, std:{}".format(i, x.std()))
            if torch.isnan(x.std()):
                print("output is nan in {} layers".format(i))
                break

        return x

    def initialize(self):
        for m in self.modules():
            if isinstance(m, nn.Linear):
                #kaiming初始化手动
                nn.init.normal_(m.weight.data, std=np.sqrt(2 / self.neural_num))

                #kaiming初始化
                # nn.init.kaiming_normal_(m.weight.data)

# flag = 0
flag = 1

if flag:
    layer_nums = 100
    neural_nums = 256
    batch_size = 16

    net = MLP(neural_nums, layer_nums)
    net.initialize()

    inputs = torch.randn((batch_size, neural_nums))  # normal: mean=0, std=1

    output = net(inputs)
    print(output)

.4. kaiming 正态分布

torch.nn.init.kaiming_normal_(tensor, a=0, mode='fan_in', nonlinearity='leaky_relu')

此为0均值的正态分布，N～ (0,std)，其中std = sqrt(2/(1+a^2)*fan_in)

其中，a为激活函数的负半轴的斜率，relu是0

mode- 可选为fan_in 或 fan_out, fan_in使正向传播时，方差一致;fan_out使反向传播时，方差一致

nonlinearity- 可选 relu 和 leaky_relu ，默认值为 。 leaky_relu

nn.init.kaiming_normal_(w, mode='fan_out', nonlinearity='relu')

.5. 均匀初始化分布

torch.nn.init.uniform_(tensor, a=0, b=1)

使值服从均匀分布U(a,b)

.6. 正态初始化分布

torch.nn.init.normal_(tensor, mean=0, std=1)

使值服从正态分布N(mean, std)，默认值为0，1

.7. 常数初始化

torch.nn.init.constant_(tensor, val)

使值为常数val nn.init.constant_(w, 0.3)

.8. 单位矩阵初始化

torch.nn.init.eye_(tensor)

将二维tensor初始化为单位矩阵（the identity matrix）

.9. 正交初始化

torch.nn.init.orthogonal_(tensor, gain=1)

使得tensor是正交的，论文:Exact solutions to the nonlinear dynamics of learning in deep linear neural networks” - Saxe, A. et al. (2013)

.10. 稀疏初始化

torch.nn.init.sparse_(tensor, sparsity, std=0.01)

从正态分布N～（0. std）中进行稀疏化，使每一个column有一部分为0

sparsity- 每一个column稀疏的比例，即为0的比例

nn.init.sparse_(w, sparsity=0.1)

注意 model.modules()和 model.children()的区别：model.modules()会迭代地遍历模型的所有子层，而model.children()只会遍历模型下的一层。

• 对网络中某一层进行初始化

self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)
init.xavier_uniform(self.conv1.weight)
init.constant(self.conv1.bias, 0.1)

• 对网络整体进行初始化

def weights_init(m):
    classname=m.__class__.__name__
    if classname.find('Conv') != -1:
        xavier(m.weight.data)
        xavier(m.bias.data)
net = Net()#构建网络
net.apply(weights_init) #apply函数会递归地搜索网络内的所有module并把参数表示的函数应用到所有的module上。
 #对所有的Conv层都初始化权重.

• 权重初始化

# Common practise for initialization.
for layer in model.modules():
    if isinstance(layer, torch.nn.Conv2d):
        torch.nn.init.kaiming_normal_(layer.weight, mode='fan_out',
                                      nonlinearity='relu')
        if layer.bias is not None:
            torch.nn.init.constant_(layer.bias, val=0.0)
    elif isinstance(layer, torch.nn.BatchNorm2d):
        torch.nn.init.constant_(layer.weight, val=1.0)
        torch.nn.init.constant_(layer.bias, val=0.0)
    elif isinstance(layer, torch.nn.Linear):
        torch.nn.init.xavier_normal_(layer.weight)
        if layer.bias is not None:
            torch.nn.init.constant_(layer.bias, val=0.0)
# Initialization with given tensor.
layer.weight = torch.nn.Parameter(tensor)

• 对指定层进行Finetune

count = 0para_optim = []for k in model.children():    count += 1    # 6 should be changed properly    if count > 6:        for param in k.parameters():            para_optim.append(param)            else:                for param in k.parameters():                    param.requires_grad = Falseoptimizer = optim.RMSprop(para_optim, lr)

• 对固定部分参数训练

# 只有True的才训练optimizer.SGD(filter(lambda p: p.requires_grad, model.parameters()), lr=1e-3)class Net(nn.Module):    def __init__(self):        super(Net, self).__init__()        self.conv1 = nn.Conv2d(1, 6, 5)        self.conv2 = nn.Conv2d(6, 16, 5)		#前面的参数就是False，而后面的不变        for p in self.parameters():            p.requires_grad=False        self.fc1 = nn.Linear(16 * 5 * 5, 120)        self.fc2 = nn.Linear(120, 84)        self.fc3 = nn.Linear(84, 10)

• 优化

optimizer = optim.Adam([        {'params': [param for name, param in net.named_parameters() if name[-4:] == 'bias'],         'lr': 2 * args['lr']},        {'params': [param for name, param in net.named_parameters() if name[-4:] != 'bias'],         'lr': args['lr'], 'weight_decay': args['weight_decay']}    ], betas=(args['momentum'], 0.999))

• 加载部分权重

# 获得模型的键值keys=[]for k,v in desnet.state_dict().items():    if v.shape:        keys.append(k)    print(k,v.shape)  # 从预训练文件中加载权重state={}pretrained_dict = torch.load('/home/lulu/pytorch/Paper_Code/weights/densenet121-a639ec97.pth')for i,(k,v) in enumerate(pretrained_dict.items()):    if 'classifier' not in k:        state[keys[i]] = v# 保存权重torch.save(state,'/home/lulu/pytorch/Paper_Code/weights/densenet121.pth')

4. 构建模型

• Sequential：顺序性，各网络层之间严格按照顺序执行，常用语block构建
• ModuleList：迭代性，常用于大量重复网络构建，通过for循环实现重复构建
• ModuleDict：索引性，常用于可选择的网络层

.1. nn.Sequential

# ============================ Sequentialclass LeNetSequential(nn.Module):    def __init__(self, classes):        super(LeNetSequential, self).__init__()        self.features = nn.Sequential(            nn.Conv2d(3, 6, 5),            nn.ReLU(),            nn.MaxPool2d(kernel_size=2, stride=2),            nn.Conv2d(6, 16, 5),            nn.ReLU(),            nn.MaxPool2d(kernel_size=2, stride=2),)         self.classifier = nn.Sequential(            nn.Linear(16*5*5, 120),            nn.ReLU(),            nn.Linear(120, 84),            nn.ReLU(),            nn.Linear(84, classes),)     def forward(self, x):        x = self.features(x)        x = x.view(x.size()[0], -1)        x = self.classifier(x)        return x

.2. nn.ModuleList

功能：像python的**list**一样包装多个网络层，以迭代的方式调用网络层

• append（）：在modulelist后面添加网络层
• extend（）：拼接两个modulelist
• insert（）：在modulelist中指定位置插入网络层

class ModuleList(nn.Module):    def __init__(self):        super(ModuleList, self).__init__()        self.linears = nn.ModuleList([nn.Linear(10, 10) for i in range(20)])    def forward(self, x):        for i, linear in enumerate(self.linears):            x = linear(x)        return xnet = ModuleList()print(net)fake_data = torch.ones((10, 10))output = net(fake_data)print(output)

.3. nn.ModuleDict

功能：像python的dict一样包装多个网络层（每一个给一个key，可通过key索引网络层）

• clear（）：清空moduleDict
• items（）：返回可迭代的键值对（key-value pairs）
• keys（）：返回字典的key
• values（）：返回字典的value
• pop（）：返回一对键值，并从字典中删除

# ============================ ModuleDictclass ModuleDict(nn.Module):    def __init__(self):        super(ModuleDict, self).__init__()        self.choices = nn.ModuleDict({            'conv': nn.Conv2d(10, 10, 3),            'pool': nn.MaxPool2d(3)        })        self.activations = nn.ModuleDict({            'relu': nn.ReLU(),            'prelu': nn.PReLU()        })    def forward(self, x, choice, act):        x = self.choices[choice](x)        x = self.activations[act](x)        return xnet = ModuleDict()fake_img = torch.randn((4, 10, 32, 32))output = net(fake_img, 'conv', 'relu')#prelu输出结果有负值，改为relu后输出没有负数，可以检查是不是按照我们的想法运行的print(output)

5. 使用预训练模型

.0. AlexNet 预训练模型修改

1. 直接使用AlexNet，并添加可视化

import osimport torchimport torch.nnimport torchvision.models as modelsimport torchvision.transforms as transformsimport torch.nn.functional as F import torchvision.utils as utilsimport cv2 import matplotlib.pyplot as pltimport numpy as np from PIL import Imageimport argparse"""input commands"""paser = argparse.ArgumentParser()paser.add_argument("--test_img", type=str, default='whippet.jpg', help="testing image")opt = paser.parse_args()# function for visualizing the feature mapsdef visualize_activation_maps(input, model):    I = utils.make_grid(input, nrow=1, normalize=True, scale_each=True)    img = I.permute((1, 2, 0)).cpu().numpy()    conv_results = []    x = input    for idx, operation in enumerate(model.features):        x = operation(x)        if idx in {1, 4, 7, 9, 11}:            conv_results.append(x)        for i in range(5):        conv_result = conv_results[i]        N, C, H, W = conv_result.size()        mean_acti_map = torch.mean(conv_result, 1, True)        mean_acti_map = F.interpolate(mean_acti_map, size=[224,224], mode='bilinear', align_corners=False)        map_grid = utils.make_grid(mean_acti_map, nrow=1, normalize=True, scale_each=True)        map_grid = map_grid.permute((1, 2, 0)).mul(255).byte().cpu().numpy()        map_grid = cv2.applyColorMap(map_grid, cv2.COLORMAP_JET)        map_grid = cv2.cvtColor(map_grid, cv2.COLOR_BGR2RGB)        map_grid = np.float32(map_grid) / 255        visual_acti_map = 0.6 * img + 0.4 * map_grid        tensor_visual_acti_map = torch.from_numpy(visual_acti_map).permute(2, 0, 1)        file_name_visual_acti_map = 'conv{}_activation_map.jpg'.format(i+1)        utils.save_image(tensor_visual_acti_map, file_name_visual_acti_map)    return 0# main if __name__ == "__main__":    """    data transforms, for pre-processing the input testing image before feeding into the net    """    data_transforms = transforms.Compose([        transforms.Resize((224,224)),             # resize the input to 224x224        transforms.ToTensor(),              # put the input to tensor format        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])  # normalize the input        # the normalization is based on images from ImageNet    ])    # obtain the file path of the testing image    test_image_dir = './alexnet_images'    test_image_filepath = os.path.join(test_image_dir, opt.test_img)    #print(test_image_filepath)    # open the testing image    img = Image.open(test_image_filepath)    print("original image's shape: " + str(img.size))    # pre-process the input    transformed_img = data_transforms(img)    print("transformed image's shape: " + str(transformed_img.shape))    # form a batch with only one image    batch_img = torch.unsqueeze(transformed_img, 0)  #Returns a new tensor with a dimension of size one inserted at the specified position.    print("image batch's shape: " + str(batch_img.shape))    # load pre-trained AlexNet model    print("\nfeed the input into the pre-trained alexnet to get the output")    alexnet = models.alexnet(pretrained=True)    # put the model to eval mode for testing    alexnet.eval()    # obtain the output of the model    output = alexnet(batch_img)    print("output vector's shape: " + str(output.shape))        # obtain the activation maps    visualize_activation_maps(batch_img, alexnet)    # map the class no. to the corresponding label    with open('class_names_ImageNet.txt') as labels:        classes = [i.strip() for i in labels.readlines()]        # print the first 5 classes to see the labels    print("\nprint the first 5 classes to see the lables")    for i in range(5):        print("class " + str(i) + ": " + str(classes[i]))        # sort the probability vector in descending order    sorted, indices = torch.sort(output, descending=True)    percentage = F.softmax(output, dim=1)[0] * 100.0    # obtain the first 5 classes (with the highest probability) the input belongs to    results = [(classes[i], percentage[i].item()) for i in indices[0][:5]]    print("\nprint the first 5 classes the testing image belongs to")    for i in range(5):        print('{}: {:.4f}%'.format(results[i][0], results[i][1]))

2. 修改Alexnet最后一层

import torchvision.models as modelsmodel = models.AlexNet()print(model)#修改网络的第一个卷积层的输入为4通道，输出的结果预测为10个类别model.features[0]=nn.Conv2d(4, 64, kernel_size=(11, 11), stride=(4, 4), padding=(2, 2))model.classifier[6] = nn.Linear(4096,10)print(model)

model = cifar10_cnn.CIFAR10_Nettest()pretrained_dict = torch.load('models/cifar10_statedict.pkl')model_dict = model.state_dict()pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}model_dict.update(pretrained_dict)model.load_state_dict(model_dict)print(model)new_model_dict = model.state_dict()dict_name = list(new_model_dict)for i, p in enumerate(dict_name):    print(i, p)print('before change:\n',new_model_dict['classifier.5.bias'])model.classifier[5]=nn.Linear(1024,17)change_model_dict = model.state_dict()new_dict_name = list(change_model_dict)print('after change:\n',change_model_dict['classifier.5.bias'])

import torch.nn as nnfrom torchvision import modelsclass BuildAlexNet(nn.Module):    def __init__(self, model_type, n_output):        super(BuildAlexNet, self).__init__()        self.model_type = model_type        if model_type == 'pre':            model = models.alexnet(pretrained=True)            self.features = model.features            fc1 = nn.Linear(9216, 4096)            fc1.bias = model.classifier[1].bias            fc1.weight = model.classifier[1].weight                        fc2 = nn.Linear(4096, 4096)            fc2.bias = model.classifier[4].bias            fc2.weight = model.classifier[4].weight                        self.classifier = nn.Sequential(                    nn.Dropout(),                    fc1,                    nn.ReLU(inplace=True),                    nn.Dropout(),                    fc2,                    nn.ReLU(inplace=True),                    nn.Linear(4096, n_output))              # #            model.classifier[6]==nn.Linear(4096,n_output)#            self.classifier = model.classifier        if model_type == 'new':            self.features = nn.Sequential(                    nn.Conv2d(3, 64, 11, 4, 2),                    nn.ReLU(inplace = True),                    nn.MaxPool2d(3, 2, 0),                    nn.Conv2d(64, 192, 5, 1, 2),                    nn.ReLU(inplace=True),                    nn.MaxPool2d(3, 2, 0),                    nn.Conv2d(192, 384, 3, 1, 1),                    nn.ReLU(inplace = True),                    nn.Conv2d(384, 256, 3, 1, 1),                    nn.ReLU(inplace=True),                    nn.MaxPool2d(3, 2, 0))            self.classifier = nn.Sequential(                    nn.Dropout(),                    nn.Linear(9216, 4096),                    nn.ReLU(inplace=True),                    nn.Dropout(),                    nn.Linear(4096, 4096),                    nn.ReLU(inplace=True),                    nn.Linear(4096, n_output))                def forward(self, x):        x = self.features(x)        x = x.view(x.size(0), -1)        out  = self.classifier(x)        return out

import numpy as npfrom torch.autograd import Variableimport torchif __name__ == '__main__':    model_type = 'pre'    n_output = 10    alexnet = BuildAlexNet(model_type, n_output)    print(alexnet)        x = np.random.rand(1,3,224,224)    x = x.astype(np.float32)    x_ts = torch.from_numpy(x)    x_in = Variable(x_ts)    y = alexnet(x_in)

.1. ResNet参数修改

resnet网络最后一层分类层fc是对1000种类型进行划分，对于自己的数据集，如果只有9类

# coding=UTF-8import torchvision.models as models#调用模型model = models.resnet50(pretrained=True)#提取fc层中固定的参数fc_features = model.fc.in_features#修改类别为9model.fc = nn.Linear(fc_features, 9)

.2. 增减卷积层

1、先建立好自己的网络（与预训练的模型类似，要不谈何fine-tune）

2、然后将预训练模型参数与自己搭建的网络不一致的部分参数去掉

3、将保留的合适的参数读入网络初始化，实现fine-tune的效果

# -*- coding:utf-8 -*-######################建立自己的网络模型net########################然后读出预训练模型参数以resnet152为例，我不是利用程序下载的，我是习惯了下载好存储在文件夹中pretrained_dict = torch.load(save_path)model_dict = net.state_dict()   #(读出搭建的网络的参数，以便后边更新之后初始化）####去除不属于model_dict的键值pretrained_dict={ k : v for k, v in pretrained_dict.items() if k in model_dict}###更新现有的model_dict的值model_dict.update(pretrained_dict)##加载模型需要的参数net.load_state_dict(model_dict)

.3. ImageNet计算多层卷积特征

class FeatureExtractor(torch.nn.Module):    """Helper class to extract several convolution features from the given    pre-trained model.    Attributes:        _model, torch.nn.Module.        _layers_to_extract, list<str> or set<str>    Example:        >>> model = torchvision.models.resnet152(pretrained=True)        >>> model = torch.nn.Sequential(collections.OrderedDict(                list(model.named_children())[:-1]))        >>> conv_representation = FeatureExtractor(                pretrained_model=model,                layers_to_extract={'layer1', 'layer2', 'layer3', 'layer4'})(image)    """    def __init__(self, pretrained_model, layers_to_extract):        torch.nn.Module.__init__(self)        self._model = pretrained_model        self._model.eval()        self._layers_to_extract = set(layers_to_extract)        def forward(self, x):        with torch.no_grad():            conv_representation = []            for name, layer in self._model.named_children():                x = layer(x)                if name in self._layers_to_extract:                    conv_representation.append(x)            return conv_representation

4. 训练特定层，冻结其它层

将模型起始的一些层的权重保持不变，重新训练后面的层，得到新的权重。在这个过程中，可多次进行尝试，从而能够依据结果找到 frozen layers 和 retrain layers 之间的最佳搭配。 如何使用预训练模型，是由数据集大小和新旧数据集(预训练的数据集和自己要解决的数据集)之间数据的相似度来决定的。

• requires_grad为False来冻结网络参数
• filter(lambda p: p.requires_grad, model.parameters())过滤掉requires_grad=false的层

#首先自己新定义一个网络class CNN(nn.Module):    def __init__(self, block, layers, num_classes=9):         #自己新定义的CNN与继承的ResNet网络结构大体相同，即除了新增层，其他层的层名与ResNet的相同。        self.inplanes = 64         super(ResNet, self).__init__() #继承ResNet网络结构        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)         self.bn1 = nn.BatchNorm2d(64)         self.relu = nn.ReLU(inplace=True)         self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)         self.layer1 = self._make_layer(block, 64, layers[0])         self.layer2 = self._make_layer(block, 128, layers[1], stride=2)         self.layer3 = self._make_layer(block, 256, layers[2], stride=2)         self.layer4 = self._make_layer(block, 512, layers[3], stride=2)         self.avgpool = nn.AvgPool2d(7, stride=1)        #新增一个反卷积层         self.convtranspose1 = nn.ConvTranspose2d(2048, 2048, kernel_size=3, stride=1, padding=1, output_padding=0, groups=1, bias=False, dilation=1)         #新增一个最大池化层         self.maxpool2 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)         #将原来的fc层改成fclass层         self.fclass = nn.Linear(2048, num_classes) #原来的fc层：self.fc = nn.Linear(512 * block.expansion, num_classes)        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))                 elif isinstance(m, nn.BatchNorm2d):                     m.weight.data.fill_(1)                     m.bias.data.zero_()                     def _make_layer(self, block, planes, blocks, stride=1):                         downsample = None                         if stride != 1 or self.inplanes != planes * block.expansion:                             downsample = nn.Sequential(                                 nn.Conv2d(self.inplanes, planes * block.expansion,                                           kernel_size=1, stride=stride, bias=False),                                 nn.BatchNorm2d(planes * block.expansion),                             )                             layers = [ ]                             layers.append(block(self.inplanes, planes, stride, downsample))                             self.inplanes = planes * block.expansion                             for i in range(1, blocks):                                 layers.append(block(self.inplanes, planes))                                 return nn.Sequential(*layers)                             def forward(self, x):                                 x = self.conv1(x)                                 x = self.bn1(x)                                 x = self.relu(x)                                 x = self.maxpool(x)                                 x = self.layer1(x)                                 x = self.layer2(x)                                 x = self.layer3(x)                                 x = self.layer4(x)                                 x = self.avgpool(x)                                 #3个新加层的forward                                 x = x.view(x.size(0), -1)                                 #因为接下来的self.convtranspose1层的输入通道是2048                                x = self.convtranspose1(x)                                 x = self.maxpool2(x)                                 x = x.view(x.size(0), -1)                                  #因为接下来的self.fclass层的输入通道是2048                                 x = self.fclass(x)                                 return x                            #加载model                             resnet50 = models.resnet50(pretrained=True)                             cnn = CNN(Bottleneck, [3, 4, 6, 3]) #创建一个自己新定义的网络对象cnn。

#微调全连接层model = torchvision.models.resnet18(pretrained=True)for param in model.parameters():    param.requires_grad = Falsemodel.fc = nn.Linear(512, 100)  # Replace the last fc layeroptimizer = torch.optim.SGD(model.fc.parameters(), lr=1e-2, momentum=0.9, weight_decay=1e-4)#以较大的学习率微调全连接层，较小的学习率微调卷积层model = torchvision.models.resnet18(pretrained=True)finetuned_parameters = list(map(id, model.fc.parameters()))conv_parameters = (p for p in model.parameters() if id(p) not in finetuned_parameters)parameters = [{'params': conv_parameters, 'lr': 1e-3},               {'params': model.fc.parameters()}]optimizer = torch.optim.SGD(parameters, lr=1e-2, momentum=0.9, weight_decay=1e-4)

.5. keypointrcnn_resnet50_fpn 模型使用

import torchimport torchvisionimport torch.nn as nndef get_model(num_kpts,train_kptHead=False,train_fpn=True):    is_available = torch.cuda.is_available()    device =torch.device('cuda:0' if is_available else 'cpu')    dtype = torch.cuda.FloatTensor if is_available else torch.FloatTensor    model = torchvision.models.detection.keypointrcnn_resnet50_fpn(pretrained=True)        for i,param in enumerate(model.parameters()):        param.requires_grad = False            if train_kptHead!=False:      for i, param in enumerate(model.roi_heads.keypoint_head.parameters()):          if i/2>=model.roi_heads.keypoint_head.__len__()/2-train_kptHead:            param.requires_grad = True    if train_fpn==True:      for param in model.backbone.fpn.parameters():        param.requires_grad = True    out = nn.ConvTranspose2d(512, num_kpts, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))    model.roi_heads.keypoint_predictor.kps_score_lowres = out        return model, device, dtype#model, device, dtype=get_model(2)