Qfyilyi

I am trying to build an image classifier using Keras 2.2.0 and tensorflow 1.9.0

I am getting an error of this sort:

str(data_shape))

ValueError: Error when checking input: expected conv2d_1_input to have shape (1, 224, 224) but got array with shape (224, 224, 3)

Here is the code:

train_datagen=ImageDataGenerator(rotation_range=40,

            width_shift_range=0.2,

            height_shift_range=0.2,

            rescale=1./255,

            shear_range=0.2,

            zoom_range=0.2,

            horizontal_flip=True,

            fill_mode='nearest')



validation_datagen=ImageDataGenerator(rescale=1./255)



train_generator = train_datagen.flow_from_directory('/media/centura/DANISH/mobile backup/moles/train/',class_mode='binary',target_size=(224, 224),batch_size=batch_size)



validation_generator = validation_datagen.flow_from_directory('/media/centura/DANISH/mobile backup/moles/test/',class_mode='binary',target_size=(224, 224),batch_size=batch_size)



#Data Dimensions

img_rows,img_cols=224,224



input_shape1=(1,img_rows,img_cols)



#initialising the model



model=Sequential()



#layer 1

model.add(Conv2D(filters=32, kernel_size=(3,3), strides=(1, 1), padding='same',input_shape=input_shape1,data_format="channels_last"))

model.add(BatchNormalization())

model.add(Activation('relu'))

#model.add(AveragePooling2D(pool_size=(2,2)))

model.add(Dropout(0.25))





#fully connected first layer



model.add(Flatten())

model.add(Dense(500,use_bias=False))

model.add(BatchNormalization())

model.add(Activation('relu'))

model.add(Dropout(0.25))      





#Fully connected final layer

model.add(Dense(1)) 

model.add(Activation('sigmoid')) 



tensorboard=TensorBoard(log_dir='logs/{}'.format(name))





model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])

#model.summary()

model.fit_generator(train_generator,epochs=50,validation_data=validation_generator,callbacks=[tensorboard])

I believe the error is coming from the train_generator. I searched in stack overflow for similar problems. I found certain solutions but it was not working for me. How can I change the dimensions of the image if it is being called from the .flow_from_directory?

Let's break down the error step-by-step to find out what it is telling us:

Error when checking input:

So it is related to the input data and input layer of the model.

expected conv2d_1_input to have shape (1, 224, 224)

If we look at the code for the first convolution layer we see that:

Conv2D(..., input_shape=input_shape1, ...)

And the value of input_shape1 as you have defined it is (1,img_rows,img_cols) which is (1, 224, 224). But:

but got array with shape (224, 224, 3)

Which means the images generated by the train_generator have a shape of (224, 224, 3) (which is correct and expected).

As a result, we see these two shapes, the shape of generated images and the given shape to input_shape argument, must be the same. Therefore, you need to modify the value of input_shape1 as follows:

input_shape1=(img_rows, img_cols, 3)

which is exactly what a convolution layer expects as its input shape (i.e. (image_height, image_width, image_channels)).