X_train 和 y_train 匹配吗？ 卷积神经网络——无分割字符识别

Question

I am trying to train a convolutional neural.network for character recognition on images. One image contains 7 characters (0-9, AZ, blank space => 37 possibilities for characters). In total there are 646 images. These images are the X_train data.

X_train.shape
(646, 600, 1200, 1)

I also have a datasheet with all labels of the images. I one-hot-encoded these labels in order to get arrays.

y_train.shape
(646, 7, 37)

646 is the number of images, so the row dimension. 7 is the length of the label, the column dimension. 37 is the number of possible characters, the depth dimension.

I would like my convolutional neural.network to recognize the characters on the image automatically without any segmentation or spatial specifications of the characters as the authors in this paper did: https://ieeexplore.ieee.org/abstract/document/8078501

I am now wondering if X_train and y_train fit together for the purpose of my neural.network? Maybe I have to reshape the dimension of X_train or y_train, because when I am running this code:

from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D

model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3), input_shape = (600, 1200, 1), activation = "relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(64, activation = "relu"))
model.add(Dense(7, activation = "relu"))
model.add(Dense(37, activation = "softmax"))

model.compile(optimizer="rmsprop", loss="categorical_crossentropy", metrics=["accuracy"])

model.fit(X_train, y_train, batch_size = 32, epochs = 10, shuffle = True)

I got the following error:

ValueError: Error when checking target: expected dense_3 to have 2 dimensions, but got array with shape (646, 7, 37)

Thank you for all advices!

Answer 1

I redefined the.networks in order to match y output shape, I can say anything about the performances

model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3), input_shape = (600, 1200, 1), activation = "relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Lambda(lambda x: tf.expand_dims(x,1)))
model.add(Dense(64, activation = "relu"))
model.add(Dense(7, activation = "relu"))
model.add(Permute((2, 1)))
model.add(Dense(37, activation = "softmax"))

the summary:

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_6 (Conv2D)            (None, 598, 1198, 32)     320       
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 299, 599, 32)      0         
_________________________________________________________________
dropout_6 (Dropout)          (None, 299, 599, 32)      0         
_________________________________________________________________
flatten_5 (Flatten)          (None, 5731232)           0         
_________________________________________________________________
lambda_3 (Lambda)            (None, 1, 5731232)        0         
_________________________________________________________________
dense_15 (Dense)             (None, 1, 64)             366798912 
_________________________________________________________________
dense_16 (Dense)             (None, 1, 7)              455       
_________________________________________________________________
permute (Permute)            (None, 7, 1)              0         
_________________________________________________________________
dense_17 (Dense)             (None, 7, 37)             74        
=================================================================