python CNNを使って画像分類器を作りたい

前提・実現したいこと

CNNを使って画像分類器を作りたい

発生している問題・エラーメッセージ

ValueError: Input 0 is incompatible with layer conv2d_1: expected ndim=4, found ndim=1

該当のソースコード

python
1from keras.models import Sequential
2from keras.layers import Conv2D,MaxPooling2D
3from keras.layers import Activation,Dropout,Flatten,Dense
4from keras.utils import np_utils
5import keras
6import numpy as np
7
8classes=["saber","archer","lancer"]
9num_classes=len(classes)
10image_size=50
11
12def main():
13    X_train,X_test,y_train,y_test=np.load("./fate_aug.npy")
14    X_train=X_train.astype("float")/256
15    X_test=X_test.astype("float")/256
16    y_train=np_utils.to_categorical(y_train,num_classes)
17    y_test=np_utils.to_categorical(y_test,num_classes)
18
19    model=model_train(X_train,y_train)
20    model_eval(model,X_test,y_test)
21
22def model_train(X,y):
23    model=Sequential()
24    model.add(Conv2D(32,(3,3),padding="same",input_shape=X.shape[1:]))
25    model.add(Activation("relu"))
26    model.add(Conv2D(32,(3,3)))
27    model.add(Activation("relu"))
28    model.add(MaxPooling2D(pool_size=(2,2)))
29    model.add(Dropout(0.25))
30
31    model.add(Conv2D(64,(3,3),padding="same"))
32    model.add(Activation("relu"))
33    model.add(Conv2D(64,(3,3)))
34    model.add(Activation("relu"))
35    model.add(MaxPooling2D(pool_size=(2,2)))
36    model.add(Dropout(0.25))
37
38    model.add(Flatten())
39    model.add(Dense(512))
40    model.add(Activation("relu"))
41    model.add(Dropout(0.5))
42    model.add(Dense(3))
43    model.add(Activation("softmax"))
44
45    opt=keras.optimizers.rmsprop(lr=0.0001,decay=1e-6)
46    model.compile(loss="categorical_crossentropy",optimizer=opt,metrics=["accuracy"])
47    model.fit(X,y,batch_size=32,epochs=50)
48
49    model.save("./fate_cnn.h5")
50
51    return model
52
53def model_eval(model,X,y):
54    scores=model.evaluate(X,y,verbose=1)
55    print("Test Loss:",scores[0])
56    print("Test Accuracy:",scores[1])
57
58if __name__=="__main__":
59    main()
60

python
1from PIL import Image
2import os,glob
3import numpy as np
4from sklearn import model_selection
5
6
7classes=["saber","archer","lancer"]
8num_classes=len(classes)
9image_size=50
10num_testdata=100
11
12X_train=[]
13X_test=[]
14Y_train=[]
15Y_test=[]
16
17
18for index,classlabel in enumerate(classes):
19    photos_dir="./"+classlabel
20    files=glob.glob(photos_dir+"/*.jpg")
21
22    for i,file in enumerate(files):
23        if i>=200:break
24        image=Image.open(file)
25        image=image.convert("RGB")
26        image=image.resize((image_size,image_size))
27
28        data=np.asarray(image)
29
30        if i<num_testdata:
31            X_test.append(data)
32            Y_test.append(index)
33
34        else:
35            for angle in range(-20,20,5):
36                image_r=image.rorate(angle)
37                data=np.asarray(image_r)
38                X_train.append(data)
39                Y_train.append(index)
40
41                image_trans=image.transpose(Image.FLIP_LEFT_RIGHT)
42                data=np.asarray(image_trans)
43                X_train.append(data)
44                Y_train.append(index)
45
46X_train=np.array(X_train)
47X_test=np.array(X_test)
48y_train=np.array(Y_train)
49y_test=np.array(Y_test)
50
51Xy=(X_train,X_test,y_train,y_test)
52
53np.save("./fate_aug.npy",Xy)
54