ResNet18识别CelebA数据集(Pytorch实战)
2024-06-21 03:56:17
文章目录
- 1.导库
- 2.获取数据集
- 自定义数据集
- 简单查看
- 创建数据集
- 3.创建Resnet18模型
- 简单查看一下网络结构
- 4.开启训练
- 训练结果
- 训练损失和测试损失关系图
- 训练精度和测试精度关系图
- 5.测试阶段
- 6.查看效果图
1.导库
import os
import timeimport numpy as np
import pandas as pdimport torch
import torch.nn as nn
import torch.nn.functional as Ffrom torch.utils.data import Dataset
from torch.utils.data import DataLoaderfrom torchvision import datasets
from torchvision import transformsimport matplotlib.pyplot as plt
from PIL import ImageDEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
2.获取数据集
http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
df1 = pd.read_csv('../input/celeba-dataset/list_attr_celeba.csv',usecols=['Male','image_id'] )
#原先数据中-1表示女性。为了方便，我们将-1改为0，即女性用数字0表示
df1.loc[df1['Male']==-1,'Male']=0
# df1.index = df1['image_id']
# del df1['image_id']
df1.head()
t = df1['Male'].values
print(t)
[0 0 1 ... 1 0 0]
#分3类，用于train,test,valid
df2 = pd.read_csv('../input/celeba-dataset/list_eval_partition.csv')df2.head(-5)
df = pd.merge(df1,df2,on='image_id')df.head()
df = df.set_index('image_id')
df.head()
df.to_csv('celeba-gender-partitions.csv')
tmp = pd.read_csv('./celeba-gender-partitions.csv', index_col=0)
tmp.head()
df.loc[df['partition'] == 0].to_csv('celeba-gender-train.csv')
df.loc[df['partition'] == 1].to_csv('celeba-gender-valid.csv')
df.loc[df['partition'] == 2].to_csv('celeba-gender-test.csv')
t1 = pd.read_csv('celeba-gender-train.csv')
t1.head()
自定义数据集
class CelebaDataset(Dataset):"""Custom Dataset for loading CelebA face images"""def __init__(self, csv_path, img_dir, transform=None):df = pd.read_csv(csv_path, index_col=0)#index_col=0 ：表示将第一列设置为index值self.img_dir = img_dir#图片所在的文件夹self.csv_path = csv_path#性别对应图片的关系self.img_names = df.index.values #such as：list_img[] = [XXXX.jpg]self.y = df['Male'].values # such as:list[]=[0 or 1]self.transform = transformdef __getitem__(self, index):img = Image.open(os.path.join(self.img_dir,self.img_names[index]))if self.transform is not None:img = self.transform(img)label = self.y[index]return img, labeldef __len__(self):return self.y.shape[0]
简单查看
tmp = CelebaDataset(csv_path = './celeba-gender-train.csv',img_dir = '../input/celeba-dataset/img_align_celeba/img_align_celeba')
x,y = tmp.__getitem__(0)
# plt.imshow(x)
print(x.size)
print(y)
(178, 218)
0
创建数据集
BATCH_SIZE = 64
custom_transform = transforms.Compose([transforms.CenterCrop((178, 178)),transforms.Resize((128, 128)),#transforms.Grayscale(), #transforms.Lambda(lambda x: x/255.),transforms.ToTensor()])train_dataset = CelebaDataset(csv_path='./celeba-gender-train.csv',img_dir='../input/celeba-dataset/img_align_celeba/img_align_celeba',transform=custom_transform)valid_dataset = CelebaDataset(csv_path='./celeba-gender-valid.csv',img_dir='../input/celeba-dataset/img_align_celeba/img_align_celeba',transform=custom_transform)test_dataset = CelebaDataset(csv_path='./celeba-gender-test.csv',img_dir='../input/celeba-dataset/img_align_celeba/img_align_celeba',transform=custom_transform)train_loader = DataLoader(dataset=train_dataset,batch_size=BATCH_SIZE,shuffle=True,num_workers=4)valid_loader = DataLoader(dataset=valid_dataset,batch_size=BATCH_SIZE,shuffle=False,num_workers=4)test_loader = DataLoader(dataset=test_dataset,batch_size=BATCH_SIZE,shuffle=False,num_workers=4)
3.创建Resnet18模型
def conv3x3(in_planes, out_planes, stride=1):"""3x3 convolution with padding"""return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,padding=1, bias=False)class BasicBlock(nn.Module):expansion = 1def __init__(self, inplanes, planes, stride=1, downsample=None):super(BasicBlock, self).__init__()self.conv1 = conv3x3(inplanes, planes, stride)self.bn1 = nn.BatchNorm2d(planes)self.relu = nn.ReLU(inplace=True)self.conv2 = conv3x3(planes, planes)self.bn2 = nn.BatchNorm2d(planes)self.downsample = downsampleself.stride = stridedef forward(self, x):residual = xout = self.conv1(x)out = self.bn1(out)out = self.relu(out)out = self.conv2(out)out = self.bn2(out)if self.downsample is not None:residual = self.downsample(x)out += residualout = self.relu(out)return outclass ResNet(nn.Module):def __init__(self, block, layers, num_classes, grayscale):self.inplanes = 64if grayscale:in_dim = 1else:in_dim = 3super(ResNet, self).__init__()self.conv1 = nn.Conv2d(in_dim, 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.AdaptiveAvgPool2d(7)self.fc = nn.Linear(25088 * block.expansion, num_classes)def _make_layer(self, block, planes, blocks, stride=1):downsample = Noneif stride != 1 or self.inplanes != planes * block.expansion:#看上面的信息是否需要卷积修改，从而满足相加条件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.expansionfor i in range(1, blocks):layers.append(block(self.inplanes, planes))return nn.Sequential(*layers)def forward(self, x):x = self.conv1(x)
# print(x.size())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)# because MNIST is already 1x1 here:# disable avg poolingx = self.avgpool(x)
# print(x.size())x = x.view(x.size(0), -1)
# print(x.size())logits = self.fc(x)
# print(x.size())probas = F.softmax(logits, dim=1)return logits, probasdef resnet18(num_classes):"""Constructs a ResNet-18 model."""model = ResNet(block=BasicBlock, layers=[2, 2, 2, 2],num_classes=num_classes,grayscale=False)return model简单查看一下网络结构
net = resnet18(num_classes = 2)
print(net)
print(net(torch.randn([1,3,256,256])))
ResNet((conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)(layer1): Sequential((0): BasicBlock((conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True))(1): BasicBlock((conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(layer2): Sequential((0): BasicBlock((conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(downsample): Sequential((0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): BasicBlock((conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(layer3): Sequential((0): BasicBlock((conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(downsample): Sequential((0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): BasicBlock((conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(layer4): Sequential((0): BasicBlock((conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(downsample): Sequential((0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): BasicBlock((conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(relu): ReLU(inplace=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(avgpool): AdaptiveAvgPool2d(output_size=7)(fc): Linear(in_features=25088, out_features=2, bias=True)
)
(tensor([[ 0.5787, -0.0013]], grad_fn=<AddmmBackward>), tensor([[0.6411, 0.3589]], grad_fn=<SoftmaxBackward>))
4.开启训练
NUM_EPOCHS = 3model = resnet18(num_classes=10)model = model.to(DEVICE)#原先这里选用SGD训练，但是效果很差，换成Adam优化就好了
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)valid_loader = test_loader
def compute_accuracy_and_loss(model, data_loader, device):correct_pred, num_examples = 0, 0cross_entropy = 0.for i, (features, targets) in enumerate(data_loader):features = features.to(device)targets = targets.to(device)logits, probas = model(features)cross_entropy += F.cross_entropy(logits, targets).item()_, predicted_labels = torch.max(probas, 1)num_examples += targets.size(0)correct_pred += (predicted_labels == targets).sum()return correct_pred.float()/num_examples * 100, cross_entropy/num_examplesstart_time = time.time()
train_acc_lst, valid_acc_lst = [], []
train_loss_lst, valid_loss_lst = [], []for epoch in range(NUM_EPOCHS):model.train()for batch_idx, (features, targets) in enumerate(train_loader):### PREPARE MINIBATCHfeatures = features.to(DEVICE)targets = targets.to(DEVICE)### FORWARD AND BACK PROPlogits, probas = model(features)cost = F.cross_entropy(logits, targets)optimizer.zero_grad()cost.backward()### UPDATE MODEL PARAMETERSoptimizer.step()### LOGGINGif not batch_idx % 500:print (f'Epoch: {epoch+1:03d}/{NUM_EPOCHS:03d} | 'f'Batch {batch_idx:04d}/{len(train_loader):04d} |' f' Cost: {cost:.4f}')# no need to build the computation graph for backprop when computing accuracymodel.eval()with torch.set_grad_enabled(False):train_acc, train_loss = compute_accuracy_and_loss(model, train_loader, device=DEVICE)valid_acc, valid_loss = compute_accuracy_and_loss(model, valid_loader, device=DEVICE)train_acc_lst.append(train_acc)valid_acc_lst.append(valid_acc)train_loss_lst.append(train_loss)valid_loss_lst.append(valid_loss)print(f'Epoch: {epoch+1:03d}/{NUM_EPOCHS:03d} Train Acc.: {train_acc:.2f}%'f' | Validation Acc.: {valid_acc:.2f}%')elapsed = (time.time() - start_time)/60print(f'Time elapsed: {elapsed:.2f} min')elapsed = (time.time() - start_time)/60
print(f'Total Training Time: {elapsed:.2f} min')训练结果
Epoch: 001/003 | Batch 0000/2544 | Cost: 0.8401
Epoch: 001/003 | Batch 0500/2544 | Cost: 0.1133
Epoch: 001/003 | Batch 1000/2544 | Cost: 0.1819
Epoch: 001/003 | Batch 1500/2544 | Cost: 0.1938
Epoch: 001/003 | Batch 2000/2544 | Cost: 0.0334
Epoch: 001/003 | Batch 2500/2544 | Cost: 0.0738
Epoch: 001/003 Train Acc.: 96.16% | Validation Acc.: 95.70%
Time elapsed: 11.19 min
Epoch: 002/003 | Batch 0000/2544 | Cost: 0.0383
Epoch: 002/003 | Batch 0500/2544 | Cost: 0.0661
Epoch: 002/003 | Batch 1000/2544 | Cost: 0.1381
Epoch: 002/003 | Batch 1500/2544 | Cost: 0.1923
Epoch: 002/003 | Batch 2000/2544 | Cost: 0.0851
Epoch: 002/003 | Batch 2500/2544 | Cost: 0.1290
Epoch: 002/003 Train Acc.: 97.19% | Validation Acc.: 96.65%
Time elapsed: 19.65 min
Epoch: 003/003 | Batch 0000/2544 | Cost: 0.0455
Epoch: 003/003 | Batch 0500/2544 | Cost: 0.0671
Epoch: 003/003 | Batch 1000/2544 | Cost: 0.0431
Epoch: 003/003 | Batch 1500/2544 | Cost: 0.0403
Epoch: 003/003 | Batch 2000/2544 | Cost: 0.1455
Epoch: 003/003 | Batch 2500/2544 | Cost: 0.1445
Epoch: 003/003 Train Acc.: 97.77% | Validation Acc.: 97.24%
Time elapsed: 28.05 min
Total Training Time: 28.05 min
训练损失和测试损失关系图
训练精度和测试精度关系图
5.测试阶段
model.eval()
with torch.set_grad_enabled(False): # save memory during inferencetest_acc, test_loss = compute_accuracy_and_loss(model, test_loader, DEVICE)print(f'Test accuracy: {test_acc:.2f}%')
Test accuracy: 97.24%
6.查看效果图
from PIL import Image
import matplotlib.pyplot as plt
for features, targets in train_loader:break
#预测环节
_, predictions = model.forward(features[:8].to(DEVICE))
predictions = torch.argmax(predictions, dim=1)
print(predictions)features = features[:7]
fig = plt.figure()# print(features[i].size())
for i in range(6):plt.subplot(2,3,i+1)plt.tight_layout()tmp = features[i]plt.imshow(np.transpose(tmp, (1, 2, 0)))plt.title("Actual value: {}".format(targets[i])+'\n'+"Prediction value: {}".format(predictions[i]),size = 10)# plt.title("Prediction value: {}".format(tname[targets[i]]))
plt.show()
ResNet18识别CelebA数据集(Pytorch实战)相关推荐
- 【Pytorch神经网络实战案例】28 GitSet模型进行步态与身份识别(CASIA-B数据集)
1 CASIA-B数据集 本例使用的是预处理后的CASIA-B数据集, 数据集下载网址如下. http://www.cbsr.ia.ac.cn/china/Gait%20Databases%20cH. ...
- Pytorch实战1：LeNet手写数字识别 (MNIST数据集)
版权说明:此文章为本人原创内容,转载请注明出处,谢谢合作! Pytorch实战1:LeNet手写数字识别 (MNIST数据集) 实验环境: Pytorch 0.4.0 torchvision 0.2. ...
- 动物数据集+动物分类识别训练代码(Pytorch)
动物数据集+动物分类识别训练代码(Pytorch) 目录 动物数据集+动物分类识别训练代码(Pytorch) 1. 前言 2. Animals-Dataset动物数据集说明 (1)Animals90动 ...
- 复盘：基于attention的多任务多模态情绪情感识别，基于BERT实现文本情感分类（pytorch实战）
复盘:基于attention机制的多任务多模态情绪情感识别(pytorch实战),基于BERT实现文本情感分类 提示:系列被面试官问的问题,我自己当时不会,所以下来自己复盘一下,认真学习和总结,以应对 ...
- Pytorch实战2：ResNet-18实现Cifar-10图像分类（测试集分类准确率95.170%）
版权说明:此文章为本人原创内容,转载请注明出处,谢谢合作! Pytorch实战2:ResNet-18实现Cifar-10图像分类 实验环境: Pytorch 0.4.0 torchvision 0.2 ...
- （！详解 Pytorch实战：①）kaggle猫狗数据集二分类：加载（集成/自定义）数据集
这系列的文章是我对Pytorch入门之后的一个总结,特别是对数据集生成加载这一块加强学习 另外,这里有一些比较常用的数据集,大家可以进行下载: 需要注意的是,本篇文章使用的PyTorch的版本是v0. ...
- celeba数据集_人脸识别常用数据集介绍（附下载链接）及常用评估指标
为什么要聊到数据集这个话题..因为数据集的noise对训练效果的影响很大!很长一段时间MegaFace的效果都上不去,就是因为数据集噪声的原因.而且自己在训练人脸的时候,如果不对数据集的噪声和属性有一 ...
- 卷积神经网络（VGG-11）识别Fashion-MNIST数据集（Pytorch版）
卷积神经网络(VGG-11)识别Fashion-MNIST数据集(Pytorch版) 1. 前言 1.1 案例介绍 1.2 环境配置 1.3 模块导入 2. 图像数据准备 2.1 训练验证集的准备 2 ...
- 【Pytorch实战4】基于CIFAR10数据集训练一个分类器
参考资料: <深度学习之pytorch实战计算机视觉> Pytorch官方教程 Pytorch中文文档 先是数据的导入与预览. import torch import torchvisio ...
最新文章
- 第二章：深入C#数据类型
- 用java解决国王的金币问题_国王赏赐金币问题（减治法）
- 软件开发生命周期中的设计阶段_BIM咨询在设计阶段包含哪些内容？体现了哪些价值？...
- 开源项目管理工具禅道ZenTaoPMS发布0.6 beta版本
- flume+elasticsearch日志收集分析
- 焦虑的移动互联网开发者如何破局？专题解析
- 断路器操作机构分合闸线圈电流特征曲线特点
- PCL—关键点检测（rangeImage）低层次点云处理
- 3d在调试区输出坐标_CSS3如何实现一个 3D 效果的魔方
- linux中Cron定时任务系统命令详解
- PyTorch 1.0 中文文档：torchvision.utils
- destoon复制新模块的方法
- 概率图模型和马尔可夫模型
- 我常用的轻量化哈希校验工具，右键菜单栏一键计算文件Hash~
- 乱得那么认真|阿里小二办公桌上的秘密~内部流出
- 用命令操作方式创建和管理数据库
- Ridge Regression and Kernel Ridge Regression
- circular 字体_DebevicCircular字体ttf下载 最新版
- 语音识别数据增强方法（google2019年7月论文）
- BZOJ 1787 AHOI2008 紧急集合 倍增LCA