I'm trying to train an autoencoder model with colored image samples but I got this error

ValueError: Dimensions must be equal, but are 476 and 480 for '{{node mean_squared_error/SquaredDifference}} = SquaredDifference[T=DT_FLOAT](model_4/conv2d_28/BiasAdd, IteratorGetNext:1)' with input shapes: [?,476,476,1], [?,480,480,3].

although i have checked the dimensions of the test and training sets all are (480,480,3)

from matplotlib import image,pyplot
import cv2


def prepro_resize(input_img):
    oimg= cv2.imread( input_img, cv2.COLOR_BGR2RGB)
    return cv2.resize(oimg, (IMG_HEIGHT, IMG_WIDTH),interpolation = cv2.INTER_AREA)

x_train_ = [(prepro_resize(x_train[i])).astype('float32')/255.0 for i in range(len(x_train))]
x_test_ = [(prepro_resize(x_test[i])).astype('float32')/255.0 for i in range(len(x_test))] 

the code for adding noise is

noise = augmenters.SaltAndPepper(0.1)
seq_object = augmenters.Sequential([noise])

train_x_n = seq_object.augment_images(x_train_ * 255) / 255
val_x_n = seq_object.augment_images(x_test_ * 255) / 255

all of the above line of codes work propertly but the previosly mentioned error occures on running model fit

history = model.fit(x_train_,train_x_n, epochs=10, batch_size=128)

i've assured that all of the samples and noise images are colored of shape(480,480,3)



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.