モデルの読み込みがうまくいきません

前提・実現したいこと

Grad-CAM++を用いて特徴量を可視化したいと考えていますが、モデルを読み込んでそのエラーが理解できずにいます
colabを用いて動かしています

発生している問題・エラーメッセージ

以下のエラーメッセージの意味が良く分かってないので、ご教示いただきたいです

AttributeError                            Traceback (most recent call last)
<ipython-input-5-b82e58f706aa> in <module>()
      7 
      8     img = img_to_array(load_img(image_path,target_size=(row,col,3)))
----> 9     img_GCAMplusplus = Grad_Cam_plus_plus(model, target_layer, img, row, col)
     10     time = time.ctime()
     11     img_Gplusplusname = image_path+time+"_GCAM++_%s.jpg"

<ipython-input-3-b1276e029972> in Grad_Cam_plus_plus(input_model, layer_name, x, row, col)
     10 
     11     # 予測クラスの算出
---> 12     predictions = model.predict(preprocessed_input)
     13     class_idx = np.argmax(predictions[0])
     14 

AttributeError: 'str' object has no attribute 'predict'

該当のソースコード

python
1import pandas as pd
2import numpy as np
3import cv2
4import argparse
5import keras
6import time
7import sys
8from keras import backend as K
9from keras.preprocessing.image import array_to_img, img_to_array, load_img
10from keras.applications.resnet import ResNet50
11
12K.set_learning_phase(1)
13
14def Grad_Cam_plus_plus(input_model, layer_name, x, row, col):
15
16    model = input_model
17
18    # 前処理
19    X = np.expand_dims(x, axis=0)
20    X = X.astype('float32')
21    preprocessed_input = X / 255.0
22
23
24    # 予測クラスの算出
25    predictions = model.predict(preprocessed_input)
26    class_idx = np.argmax(predictions[0])
27
28    #  使用する重みの抽出、高階微分の計算
29    class_output = model.layers[-1].output
30    conv_output = model.get_layer(layer_name).output
31    grads = K.gradients(class_output, conv_output)[0]
32    #first_derivative：１階微分
33    first_derivative = K.exp(class_output)[0][class_idx] * grads
34    #second_derivative：２階微分
35    second_derivative = K.exp(class_output)[0][class_idx] * grads * grads
36    #third_derivative：３階微分
37    third_derivative = K.exp(class_output)[0][class_idx] * grads * grads * grads
38
39    #関数の定義
40    gradient_function = K.function([model.input], [conv_output, first_derivative, second_derivative, third_derivative])  # model.inputを入力すると、conv_outputとgradsを出力する関数
41
42
43    conv_output, conv_first_grad, conv_second_grad, conv_third_grad = gradient_function([preprocessed_input])
44    conv_output, conv_first_grad, conv_second_grad, conv_third_grad = conv_output[0], conv_first_grad[0], conv_second_grad[0], conv_third_grad[0]
45
46    #alphaを求める
47    global_sum = np.sum(conv_output.reshape((-1, conv_first_grad.shape[2])), axis=0)
48    alpha_num = conv_second_grad
49    alpha_denom = conv_second_grad*2.0 + conv_third_grad*global_sum.reshape((1,1,conv_first_grad.shape[2]))
50    alpha_denom = np.where(alpha_denom!=0.0, alpha_denom, np.ones(alpha_denom.shape))
51    alphas = alpha_num / alpha_denom
52
53    #alphaの正規化
54    alpha_normalization_constant = np.sum(np.sum(alphas, axis = 0), axis = 0)
55    alpha_normalization_constant_processed = np.where(alpha_normalization_constant != 0.0, alpha_normalization_constant, np.ones(alpha_normalization_constant.shape))
56    alphas /= alpha_normalization_constant_processed.reshape((1,1,conv_first_grad.shape[2]))
57
58    #wの計算
59    weights = np.maximum(conv_first_grad, 0.0)
60    deep_linearization_weights = np.sum((weights * alphas).reshape((-1, conv_first_grad.shape[2])))
61
62    #Lの計算
63    grad_CAM_map = np.sum(deep_linearization_weights * conv_output, axis=2)
64    grad_CAM_map = np.maximum(grad_CAM_map, 0)
65    grad_CAM_map = grad_CAM_map / np.max(grad_CAM_map)
66
67    #ヒートマップを描く
68    grad_CAM_map = cv2.resize(grad_CAM_map, (row, col), cv2.INTER_LINEAR)
69    jetcam = cv2.applyColorMap(np.uint8(255 * grad_CAM_map), cv2.COLORMAP_JET)  # モノクロ画像に疑似的に色をつける
70    jetcam = (np.float32(jetcam) + x / 2)   # もとの画像に合成
71
72    return jetcam
73
74if __name__ == '__main__':
75    model = "/content/drive/MyDrive/ji/saved_model/my_model"
76    target_layer = 'batch_normalization'
77    image_path = '/content/drive/MyDrive/Cs-N013.jpg'
78    row = 224
79    col = 224
80
81    img = img_to_array(load_img(image_path,target_size=(row,col,3)))
82    img_GCAMplusplus = Grad_Cam_plus_plus(model, target_layer, img, row, col)
83    time = time.ctime()
84    img_Gplusplusname = image_path+time+"_GCAM++_%s.jpg"
85    cv2.imwrite(img_Gplusplusname, img_GCAMplusplus)
86    print("Completed.")