混淆矩阵的绘制

提示：文章写完后，目录可以自动生成，如何生成可参考右边的帮助文档

使用Python绘制混淆矩阵

• 前言
• 具体步骤
• • 1.引入库
  • 2.设置参数
  • 3.混淆矩阵定义
  • 4.计算准确率及绘制混淆矩阵
• 绘制结果

---

前言

主要展示在分类算法预测的过程中，加入混淆矩阵的绘制。

---

具体步骤 1.引入库

代码如下（示例）：

import argparse

import torch
from torch.backends import cudnn
from torchvision import transforms
from torch.utils.data import DataLoader, Dataset
import torch.nn.functional as F
import torch.nn as nn
import numpy as np

from data_loaders import Plain_Dataset, eval_data_dataloader
from model import ResidualNet  # 引入模型

import matplotlib.pyplot as plt

2.设置参数

代码如下（示例）：

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

parser = argparse.ArgumentParser(description="Configuration of testing process")
parser.add_argument('-m', '--model', type=str,default='./model/RestNet18.pt')
parser.add_argument('-depth', default=18, type=int)
parser.add_argument('-d', '--data', type=str, default='')
parser.add_argument('-att_type', default='se', choices=['cbam', 'se'], type=str)
args = parser.parse_args()

transformation = transforms.Compose([transforms.ToTensor(),transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))])
test_path = args.data + '/' + 'test'
dataset = Plain_Dataset( img_dir=test_path, datatype='test',transform=transformation)
test_loader =  DataLoader(dataset,batch_size=64,num_workers=0)

# 加载模型
net = ResidualNet('CIFAR10', args.depth, 7, args.att_type)
net.load_state_dict(torch.load(args.model))
net.to(device)

3.混淆矩阵定义

代码如下（示例）：

# 混淆矩阵定义
def confusion_matrix(preds,labels,conf_matrix):
    for p,t in zip(preds,labels):
        conf_matrix[p,t] += 1
    return conf_matrix

def plot_maxtrix(maxtrix,per_kinds):
 	# 分类标签
    lables = ['Angry', 'Disgust', 'Fear', 'Happy', 'Neutral', 'Sad', 'Surprise']
    
    Maxt = np.empty(shape=[0,7])

    m = 0
    for i in range(7):
        print('row sum:',per_kinds[m])
        f = (maxtrix[m,:]*100)/per_kinds[m]
        Maxt = np.vstack((Maxt,f))
        m = m+1

    thresh = Maxt.max()/1

    plt.imshow(Maxt, cmap=plt.cm.Blues)

    for x in range(7):
        for y in range(7):
            info = float(format('%.1f' % F[y,x]))
            print('info:',info)
            plt.text(x,y,info,verticalalignment='center',horizontalalignment='center')
    plt.tight_layout()
    plt.yticks(range(7),lables)  # y轴标签
    plt.xticks(range(7),lables,rotation=45)  # x轴标签
    plt.savefig('./test.png',bbox_inches='tight')  # bbox_inches='tight'可确保标签信息显示全
    plt.show()

4.计算准确率及绘制混淆矩阵

代码如下（示例）：

if __name__ == '__main__':
	with torch.no_grad():
        for data, labels in test_loader:
            data, labels = data.to(device), labels.to(device)

            outputs = net(data)
            pred = F.softmax(outputs,dim=1)
            classs = torch.argmax(pred,1)

            conf_maxtri = confusion_matrix(classs,labels,conf_maxtri)
            conf_maxtri = conf_maxtri.cpu()

            wrong = torch.where(classs != labels,torch.tensor([1.]).cuda(),torch.tensor([0.]).cuda())
            acc = 1- (torch.sum(wrong) / 64)  # 64为batch size
            total.append(acc.item())

    print('测试集的准确率为: %f %%' % (100 * np.mean(total)))
   
    # 绘制混淆矩阵
    conf_maxtri = np.array(conf_maxtri.cpu())
    corrects = conf_maxtri.diagonal(offset=0)
    per_kinds = conf_maxtri.sum(axis=1)
    plot_maxtrix(conf_maxtri,per_kinds)

---

绘制结果