test_model_up.py

# -*- coding: utf-8 -*-
"""
Created on Thu Jan 23 22:47:47 2020

@author: Leo
"""

# test
import time
import numpy as np
import matplotlib.pyplot as plt
from keras.models import load_model
from keras.models import model_from_json
from keras.preprocessing import image


def array_img_reshape(array, target_shape):
    import PIL
    # PIL.Image.BILINEAR
    array = image.img_to_array(image.array_to_img(array).resize(target_shape, resample=PIL.Image.BILINEAR))
    return array

def preprocess_img(img_id,target_shape):
    '''
    加载图片，切分，归一化
    输出：子图集，最后一张子图的高度
    '''
    img = image.load_img(img_id)
    img = image.img_to_array(img)[...,:1]
    h, _, _ = np.shape(img)
    n = int(np.ceil(h/target_shape[0]))
    imgs = np.zeros((n, *target_shape, 1))
    for i in range(n):
        if i != n-1:
            imgs[i,...] = img[target_shape[0]*i:target_shape[0]*(i+1),...]
        else:
            imgs[i,...] = array_img_reshape(img[target_shape[0]*i:], target_shape)
    imgs = imgs/255.0
    return imgs, h-(n-1)*target_shape[0]

def plot_result(sort_list_i, result, delta_h, target_size):
    img1  = image.load_img(sort_list_i[1]) # src
    img2  = image.load_img(sort_list_i[2]) # label
    imgs1 = np.zeros_like(img2)[...,:1]    # output1
    imgs2 = np.zeros_like(img2)[...,:1]    # output2
    for k in range(np.shape(result[0])[0]):
        fig_k = result[0][k,...,:1]
        if k == np.shape(result[0])[0]-1:
            imgs1[target_size[0]*k:,...] = array_img_reshape(fig_k, (target_size[0], delta_h))
        else:
            imgs1[target_size[0]*k:target_size[0]*(k+1),...] = array_img_reshape(fig_k, target_size)
    for j in range(np.shape(result[0])[0]):
        fig_j = result[0][j,...,1:]
        if j == np.shape(result[0])[0]-1:
            imgs2[target_size[0]*j:,...] = array_img_reshape(fig_j, (target_size[0], delta_h))
        else:
            imgs2[target_size[0]*j:target_size[0]*(j+1),...] = array_img_reshape(fig_j, target_size)
    imgs2 = np.concatenate((imgs1,imgs2),axis=2)
    imgs2 = np.reshape(imgs2,(1,-1,2))
    imgs2 = np.argmax(imgs2,axis=2).astype(np.int8)
    imgs2 = np.reshape(imgs2,np.shape(imgs1))

    plt.figure(figsize=(10,10))
    plt.subplot(141)
    plt.imshow(img1)
    plt.title('src')
    plt.axis('off')
    plt.subplot(142)
    plt.imshow(img2)
    plt.title('label')
    plt.axis('off')
    plt.subplot(143)
    plt.imshow(255-imgs1[...,0],cmap='gray')
    plt.title('output')
    plt.axis('off')
    plt.subplot(144)
    plt.imshow(imgs2[...,0],cmap='gray')
    plt.title('result')
    plt.axis('off')
    #plt.show()
    return None

def load_model_from_json(model_path, model_name):
    with open(model_path+model_name+'.json', 'r') as file:
        model_json = file.read()
    model = model_from_json(model_json)
    model.load_weights(model_path+model_name+'_weights.h5')
    return model

if __name__ == '__main__':
    # 加载测试数据
    import data_manager
    #test_set_path  = './KolektorSDD/test_set'
    test_set_path  = './KolektorSDD/all'
    sort_list = data_manager.data_manager(test_set_path)
    # 模型的输入形状
    target_size = (500, 500)
    output_size = (500, 500)
    n_label = 2
    # 加载模型 / 权重
    model_path = './model/'
    model_name = 'best_model_up_500'

    try:
        '''
        两者选其一
        1. 直接加载模型
        2. 加载json文件和weights权重
        '''
        #model = load_model(model_path+model_name+'.h5')     # 1.
        model = load_model_from_json(model_path, model_name) # 2.
        print('Model weights loaded!')
    except:
        raise FileNotFoundError('Model not found!')
    
    model.summary()
    TP, FP, FN, TN = 0, 0, 0, 0
    for i in range(len(sort_list)):
        print('开始处理第',str(i+1),'张图片...')
        t1 = time.time()
        img_set, delta_h = preprocess_img(sort_list[i][1],target_size)
        result = model.predict(img_set, verbose=0)
        # 判断所有子图中是否包含class为 1 的输出
        class_i = 0
        for j in range(np.shape(result[0])[0]):
            class_i = np.max((class_i, int(list(np.rint(result[1][j])).index(1)), 0))
        t2 = time.time()
        if class_i != sort_list[i][0]:
            print('第{}例分类错误！耗时：{:.4f}s'.format(i+1, t2-t1))
            # 画出分类错误的实例
            plot_result(sort_list[i], result, delta_h, target_size)
        else:
            print('第{}例分类正确！耗时：{:.4f}s'.format(i+1, t2-t1))
            # 画出分类正确的正例
            if class_i == 1:
                plot_result(sort_list[i], result, delta_h, target_size)
        # 统计 TP, FP, FN, TN
        if class_i == 1:                 # P
            if sort_list[i][0] == 1:     # T
                TP += 1
            elif sort_list[i][0] == 0:   # F
                FP += 1
        elif class_i == 0:               # N
            if sort_list[i][0] == 1:     # F
                FN += 1
            elif sort_list[i][0] == 0:   # T
                TN += 1
    print('TP：{}  FP：{}  FN：{}  TN：{}'.format(TP, FP, FN, TN))
    print('准确率：%.3f%%'%((TP+TN)/(TP+TN+FP+FN)*100))
    print('查准率：%.3f%%'%(TP/(TP+FP)*100))
    print('查全率：%.3f%%'%(TP/(TP+FN)*100))