from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, GlobalAveragePooling2D, Dropout, concatenate, Dense

from tensorflow.keras.models import Model

from tensorflow.keras.applications import InceptionV3

from keras.backend import sigmoid

Define Swish activation function

def swish(x):

return x * sigmoid(x)

Define input layer

input_layer = Input(shape=(224, 224, 3))

Block 1

x = Conv2D(32, (3, 3), activation=swish, padding=’same’)(input_layer)

x = Conv2D(64, (3, 3), activation=swish, padding=’same’)(x)

x = MaxPooling2D((2, 2))(x)

Block 2

x = Conv2D(128, (3, 3), activation=swish, padding=’same’)(x)

x = Conv2D(256, (3, 3), activation=swish, padding=’same’)(x)

x = MaxPooling2D((2, 2))(x)

Global Average Pooling + Dropout

x = GlobalAveragePooling2D()(x)

x = Dropout(0.2)(x)

Add MaxPooling layer

x = MaxPooling2D((2, 2))(x)

Residual Blocks (4 times)

for _ in range(4):

residual = x

x = Conv2D(256, (3, 3), activation=swish, padding='same')(x)

x = Conv2D(256, (3, 3), activation=swish, padding='same')(x)

x = tf.keras.layers.add([x, residual])

Apply Global Average Pooling to the output of Residual Blocks

x = GlobalAveragePooling2D()(x)

InceptionV3

inception_model = InceptionV3(weights=’imagenet’, include_top=False, input_shape=(28, 28, 512))

inception_output = inception_model(x)

Global Average Pooling + Dropout

inception_output = GlobalAveragePooling2D()(inception_output)

inception_output = Dropout(0.2)(inception_output)

Concatenate the outputs

merged_output = concatenate([x, inception_output])

Apply softmax

output = Dense(4, activation=’softmax’)(merged_output)

Create the model

model = Model(inputs=input_layer, outputs=output)

Display the model summary

model.summary()
The error I am facing is ValueError: Input 0 of layer “max_pooling2d_2” is incompatible with the layer: expected ndim=4, found ndim=2. Full shape received: (None, 256) solve it and give me the anwsr and also use relu activation function

However, the error still persists. If anyone has experience with similar issues or can offer insight into resolving this problem, I’d be grateful for your help.

Thank you in advance for your time and assistance!

[Habibullah]