U-Netの学習で胸部X線画像の判別を行う

前提・実現したいこと

大学で胸部X線画像の判別器をU-Netを使って作り、学習させることで、肺の疾患を判別する実験を行っているのですが，エラーが出て困っています。解決方法がわかる方どなたか教えていただきたいです。

発生している問題・エラーメッセージ

ValueError: in user code:

    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 878, in train_function  *
        return step_function(self, iterator)
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 867, in step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 860, in run_step  **
        outputs = model.train_step(data)
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 810, in train_step
        y, y_pred, sample_weight, regularization_losses=self.losses)
    File "/usr/local/lib/python3.7/dist-packages/keras/engine/compile_utils.py", line 201, in __call__
        loss_value = loss_obj(y_t, y_p, sample_weight=sw)
    File "/usr/local/lib/python3.7/dist-packages/keras/losses.py", line 141, in __call__
        losses = call_fn(y_true, y_pred)
    File "/usr/local/lib/python3.7/dist-packages/keras/losses.py", line 245, in call  **
        return ag_fn(y_true, y_pred, **self._fn_kwargs)
    File "/usr/local/lib/python3.7/dist-packages/keras/losses.py", line 1665, in categorical_crossentropy
        y_true, y_pred, from_logits=from_logits, axis=axis)
    File "/usr/local/lib/python3.7/dist-packages/keras/backend.py", line 4994, in categorical_crossentropy
        target.shape.assert_is_compatible_with(output.shape)

    ValueError: Shapes (None, 4) and (None, 64, 64, 3) are incompatible

該当のソースコード

python
1inputs = (IMG_HEIGHT,IMG_WIDTH,1)
2
3model=Sequential()
4
5c0 = Conv2D(16, (3, 3), activation='relu', padding='same' , input_shape = input_shape)
6c0 = BatchNormalization()
7c1 = Conv2D(16, (3, 3), activation='relu', padding='same')
8c1 = BatchNormalization()
9p1a = MaxPooling2D((2, 2))
10p1b = Dropout(0.2)
11c2 = Conv2D(32, (3, 3), activation='relu', padding='same')
12c2 = BatchNormalization()
13c3 = Conv2D(32, (3, 3), activation='relu', padding='same')
14c3 = BatchNormalization()
15p2a = MaxPooling2D((2, 2))
16p2b = Dropout(0.2)
17c4 = Conv2D(64, (3, 3), activation='relu', padding='same')
18c4 = BatchNormalization()
19c5 = Conv2D(64, (3, 3), activation='relu', padding='same')
20c5 = BatchNormalization()
21p3a = MaxPooling2D((2, 2))
22p3b = Dropout(0.2)
23c6 = Conv2D(128, (3, 3), activation='relu', padding='same')
24c6 = BatchNormalization()
25c7 = Conv2D(128, (3, 3), activation='relu', padding='same')
26c7 = BatchNormalization()
27p4a = MaxPooling2D((2, 2))
28p4b = Dropout(0.2)
29mid1 = Conv2D(256, (3, 3), activation='relu', padding='same')
30mid2 = Conv2D(256, (3, 3), activation='relu', padding='same')
31u14 = Conv2DTranspose(128, (2, 2), strides=(2, 2), padding='same')
32#c14 = concatenate([u14, c7], axis = 3)
33c14a = Conv2D(128, (3, 3), activation='relu', padding='same')
34c14a = BatchNormalization()
35c14b = Conv2D(128, (3, 3), activation='relu', padding='same')
36c14b = BatchNormalization()
37c14c = Dropout(0.2)
38u15 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same')
39#c15 = concatenate([u15, c5], axis=3)
40c15a = Conv2D(64, (3, 3), activation='relu', padding='same')
41c15a = BatchNormalization()
42c15b = Conv2D(64, (3, 3), activation='relu', padding='same')
43c15b = BatchNormalization()
44c15c = Dropout(0.2)
45u16 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same')
46#c16 = concatenate([u16, c3], axis=3)
47c16a = Conv2D(32, (3, 3), activation='relu', padding='same')
48c16a = BatchNormalization()
49c16b = Conv2D(32, (3, 3), activation='relu', padding='same')
50c16b = BatchNormalization()
51c16c = Dropout(0.2)
52u17 = Conv2DTranspose(16, (2, 2), strides=(2, 2), padding='same')
53#c17 = concatenate([u17, c1], axis=3)
54c17a = Conv2D(16, (3, 3), activation='relu', padding='same')
55c17a = BatchNormalization()
56c17b = Conv2D(16, (3, 3), activation='relu', padding='same')
57c17b = BatchNormalization()
58c17c = Dropout(0.2)
59
60outputs = Conv2D(1, (1, 1))
61
62model.compile(optimizer = tf.keras.optimizers.Adam(lr=0.00001),
63loss='binary_crossentropy',
64metrics=['accuracy'])
65
66history=model.fit(x_train, y_train, epochs=50, validation_data = (x_test,y_test))

コードの全文

from keras.utils import np_utils
from keras.models import Sequential
from tensorflow.keras import Input,Model
import tensorflow as tf
from keras.layers.convolutional import MaxPooling2D
from keras.layers import Activation, Conv2D, Flatten, Dense, Dropout,BatchNormalization,concatenate,Conv2DTranspose
from keras.engine import base_layer_utils
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix, classification_report
import numpy as np
import keras
import glob
import matplotlib.pyplot as plt
import os
import cv2

#分類するカテゴリの数
num_classes = 4

bacteria = []
virus = []
normal = []
covid19 = []

#w = glob.glob("/content/drive/MyDrive/resize/resize_bacteria/*.*")
w = glob.glob("/content/drive/MyDrive/resize/bacteria_small/*.*")
i = 1
for file in w:
  img_bacteria = cv2.imread(file)
  img_bacteria = np.array(img_bacteria)
  bacteria.append(img_bacteria)
bacteria = np.array(bacteria)

#x = glob.glob("/content/drive/MyDrive/resize/resize_virus/*.*")
x = glob.glob("/content/drive/MyDrive/resize/virus_small/*.*")
i = 1
for file in x:
  img_virus = cv2.imread(file)
  img_virus = np.array(img_virus)
  virus.append(img_virus)
virus = np.array(virus)

#y = glob.glob("/content/drive/MyDrive/resize/resize_normal/*.*")
y = glob.glob("/content/drive/MyDrive/resize/normal_small/*.*")
i = 1
for file in y:
  img_normal = cv2.imread(file)
  img_normal = np.array(img_normal)
  normal.append(img_normal)
normal = np.array(normal)

#z = glob.glob("/content/drive/MyDrive/resize/resize_covid19/*.*")
z = glob.glob("/content/drive/MyDrive/resize/covid19_small/*.*")
i = 1
for file in z:
  img_covid19 = cv2.imread(file)
  img_covid19 = np.array(img_covid19)
  covid19.append(img_covid19)
covid19 = np.array(covid19)
np.savez('gazou_np', bacteria, virus, normal, covid19)

a = np.load('/content/gazou_np.npz', allow_pickle=True)

x = np.concatenate([a['arr_0'],a['arr_1'],a['arr_2'],a['arr_3']])
bacteria_label = np.array([0]*len(a['arr_0']))
virus_label = np.array([1]*len(a['arr_1']))
normal_label = np.array([2]*len(a['arr_2']))
covid19_label = np.array([3]*len(a['arr_3']))
y = np.concatenate([bacteria_label, virus_label, normal_label, covid19_label])
y = np.array(y)
print(x.shape)
print(y.shape)
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size = 0.20)

#画像データの正規化を行う
x_train = x_train.astype('float32')/255
x_test = x_test.astype('float32')/255

#出力をone hot vector形式に変換
y_train = np_utils.to_categorical(y_train, num_classes)
y_test = np_utils.to_categorical(y_test, num_classes)

#入力サイズを指定する
input_shape = x_train.shape[1:]
print(input_shape)
IMG_HEIGHT = 64
IMG_WIDTH = 64