数据可视化

1. 绘制各种曲线

1.1 损失函数曲线

def plot_loss_cure(epoch_list, loss_list):
    plt.plot(epoch_list, loss_list)
    plt.xlabel('Number of Epoch')
    plt.ylabel('Loss')
    plt.title('RNN')
    plt.savefig('rnn.png')  # 保存图片
    plt.show()

epoch_list = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
loss_list = [0.1125, 0.0307, 0.0218, 0.0172, 0.0154, 0.0138, 0.011, 0.0123, 0.0106, 0.0103]
plot_loss_cure(epoch_list, loss_list)

效果如下

 1.2 准确率曲线

epoch_list = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
acc_list = [0.9076, 0.9517, 0.9077, 0.9662, 0.9694, 0.9685, 0.9765, 0.9711, 0.9744, 0.9790]

plot_loss_cure(epoch_list, acc_list)

效果如下：

1.3 同时画出训练集验证集准确率与损失曲线对比图

from collections import defaultdict

def plot_training_history(history):
    # 按1行2列格式显示
    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 4))
    ax1.plot(history['train_loss'], label='train_loss')
    ax1.plot(history['val_loss'], label='val_loss')
    ax1.legend()
    ax1.set_ylabel('Loss')
    ax1.set_xlabel('Epoch')
    # ax1.set_title('Model Loss')

    ax2.plot(history['train_acc'], label='train_acc')
    ax2.plot(history['val_acc'], label='val_acc')
    ax2.legend()
    ax2.set_ylabel('Acc')
    ax2.set_xlabel('Epoch')

    fig.suptitle('Training History')
    plt.pause(0.001)
    plt.show()

调用

history = defaultdict(list)  # 构建一个默认value为list的字典
train_acc = [0.9176, 0.9417, 0.9517, 0.9622, 0.9694, 0.9715, 0.9765, 0.9771, 0.9784, 0.9790]
val_acc = [0.8076, 0.8517, 0.8077, 0.8662, 0.8694, 0.8685, 0.8765, 0.8711, 0.8744, 0.8790]
train_loss = [0.1225, 0.0407, 0.0318, 0.0192, 0.0184, 0.0158, 0.0141, 0.0123, 0.0116, 0.0113]
val_loss = [0.2125, 0.2307, 0.2218, 0.2172, 0.1154, 0.1138, 0.111, 0.1123, 0.1106, 0.1103]
history['train_acc'].extend(train_acc)
history['val_acc'].extend(val_acc)
history['train_loss'].extend(train_loss)
history['val_loss'].extend(val_loss)
plot_training_history(history)

效果如下：

1.4 绘制ROC曲线

import plotly.offline as py
# py.init_notebook_mode(connected=True)
import plotly.graph_objs as go
import plotly.subplots as tls
import plotly.figure_factory as ff
from sklearn.metrics import roc_curve, roc_auc_score

def plot_roc_auc_curve(y_test, y_pred):
    # y_test.squeeze()  # 有时需要降维
    fpr, tpr, t = roc_curve(y_test, y_pred)
    model_roc_auc = roc_auc_score(y_test, y_pred)
    trace = go.Scatter(x=fpr, y=tpr,
                       name='ROC: ' + str(model_roc_auc),
                       line=dict(color=('rgb(22, 96, 167)'), width=2), fill='tozeroy')
    fig = tls.make_subplots(rows=1, cols=1, print_grid=False, subplot_titles=('ROC_AUC_curve',))
    fig.append_trace(trace, 1, 1)
    fig.layout.xaxis.update(dict(title='FPR'), range=[0, 1.05])
    fig.layout.yaxis.update(dict(title='TPR', range=[0, 1.05]))
    fig.layout.titlefont.size = 14
    py.plot(fig)

if __name__ == '__main__':
    y_test = [0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1]
    y_pred = [1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1]
    plot_roc_auc_curve(y_test, y_pred)

效果展示：

1.5 绘制混淆矩阵

from sklearn.metrics import confusion_matrix

def plot_cm(y_test, y_pred):
    cm = confusion_matrix(y_test, y_pred)
    trace1 = go.Heatmap(z=cm,
                        x=["0(pred)", "1(pred)"],
                        y=["0(True)", "1(True)"],
                        xgap=2, ygap=2,
                        colorscale='Viridis', showscale=False)
    TP = cm[1, 1]
    FN = cm[1, 0]
    FP = cm[0, 1]
    TN = cm[0, 0]
    accuracy = (TP+TN) / (TP+TN+FP+FN)
    precision = TP / (TP+FP)
    recall = TP / (TP+FN)
    F1_score = (2*precision*recall) / (precision+recall)
    df_cm = pd.DataFrame(data=[[accuracy, precision, recall, F1_score]])
    df_cm = df_cm.T

    # 可视化
    colors = ['gold', 'lightgreen', 'lightcoral', 'lightskyblue']
    trace2 = go.Bar(x=df_cm[0].values,
                    y=['Accuracy', 'Precision', 'Recall', 'F1_score'],
                    text=np.round_(df_cm[0].values, 4),
                    textposition='auto',
                    orientation='h',
                    opacity=0.8,
                    marker=dict(color=colors, line=dict(color='#000000', width=1.5)))
    fig = tls.make_subplots(rows=2, cols=1, subplot_titles=('Confusion Matrix', 'Metrics'))
    fig.append_trace(trace1, 1, 1)
    fig.append_trace(trace2, 2, 1)
    py.plot(fig)

if __name__ == '__main__':
    y_test = [0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1]
    y_pred = [1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1]
    plot_cm(y_test, y_pred)

效果展示：

 

2. 显示图片

 2.1 绘制一张图片

方法1：

def image_show_1(img, title=None):
    plt.figure(figsize=(14, 3))
    img = img.numpy().transpose((1, 2, 0))  # 改变shape顺序
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])
    img = img * std + mean  # 逆向复原
    img = np.clip(img, 0, 1)  # 将元素值限制在0-1之间
    plt.imshow(img)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)
    plt.show()

方法2：

def image_show_2(img):
    img = img / 2 + 0.5  # 逆正则化
    np_img = img.numpy()  # tensor --> numpy
    plt.imshow(np.transpose(np_img, (1, 2, 0)))  # 改变通道顺序
    plt.show()

方法3： 从url读取图片

from PIL import Image

def image_show_3(img):
    plt.figure(figsize=(8, 8))
    plt.imshow(img)
    plt.axis('off')   # 关闭坐标轴
    plt.show()

img = Image.open(data_loader['train'].dataset.root + '/NORMAL/IM-0117-0001.jpeg')
image_show_3(img)

2.2 以grid形式绘制一幅图片

显示一个batch_size的图片

def show_a_batch_train_data_picture(dataloaders, target_names):
    datas, targets = next(iter(dataloaders['train']))
    print(targets)  # 需要根据target_names标签列表进行映射
    # tensor([1, 0, 0, 1, 1, 0, 1, 1])

    # 将若干张图片拼成一幅图像
    grid = make_grid(datas, nrow=4, padding=10)
    # 图片显示
    image_show_1(grid, title=[target_names[x] for x in targets])
    image_show_2(grid)
    # 显示第6张图片
    image_show_1(*data_loader['train'].dataset[5])

if __name__ == "__main__":
    BATCH_SIZE = 8
    DATA_PATH = '../data/chest_xray'
    data_loader = load_data(DATA_PATH, BATCH_SIZE)
    # 获取标签的类别名称
    labels = ['NORMAL', 'PNEUMONIA']
    show_a_batch_train_data_picture(data_loader, labels)