机器学习--人脸自动补齐（11）

随机树ExtraTreeRegressor()： 分列点随机选取、不考虑信息增益，减少过拟合

获取一个对象所属的类名称： model.class.name

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

# 导入人脸数据集
from sklearn.datasets import fetch_olivetti_faces,fetch_lfw_people
faces = fetch_olivetti_faces()

images = faces.images
data = faces.data
target = faces.target

# 训练集和测试集拆分
# 训练集由每个人提供九张照片
# 测试集每个人呢提供2张照片

# 样本特征和照片
# 每张照片扁平处理，前半部分2048之前，为样本特征X
# 每张照片扁平处理，前半部分2048之后，为样本标签X
def precess_faces(images):
    # 正常图像处理步骤：1.图像定位（opencv）2.图像大小规范 3.灰度处理 4.扁平化
    X_train = []
    X_test = []
    Y_train = []
    Y_test = []
    
    idx = 0
    for i in range(40):
        for j in range(10):
            img_data = images[idx].ravel()
            upper_data = img_data[:2048]
            bottom_data = img_data[2048:]
            idx += 1
            if j < 9:
                X_train.append(upper_data)
                Y_train.append(bottom_data)
            else:
                X_test.append(upper_data)
                Y_test.append(bottom_data)
    return np.array(X_train),np.array(X_test),np.array(Y_train),np.array(Y_test)

# 拆分 
X_train,X_test,Y_train,Y_test = precess_faces(images)

# 建模预测回归
from sklearn.neighbors import KNeighborsRegressor
from sklearn.linear_model import LinearRegression,Ridge,Lasso
from sklearn.tree import DecisionTreeRegressor,ExtraTreeRegressor

knn = KNeighborsRegressor()
liner = LinearRegression()
ridge = Ridge()
ls = Lasso()
et = ExtraTreeRegressor()
dt = DecisionTreeRegressor()

# 得到预测结果
result = {}
for model in [knn,liner,ridge,ls,et,dt]:
    model.fit(X_train,Y_train)
    Y_ = model.predict(X_test)
    result[model.__class__.__name__] = Y_
    
# 将真实值添加进结果
result['True'] = Y_test

# 展示预测结果

plt.figure(figsize=(24,30))

ax_loc = 1
for idx in range(5):
    upper_face_data = X_test[idx]
    for  key,Y_ in result.items():
        bottom_face_data = Y_[idx]
        plt.subplot(5,7,ax_loc)
        ax_loc +=1
        plt.imshow(np.concatenate((upper_face_data,bottom_face_data)).reshape(64,64),plt.cm.gray)
        plt.title(key,fontsize=15,color='red')