[Python] Issue with Input Shape Compatibility for Pre-trained VGG16 Model in Keras

I came across a paper where they use a sub-model to extract features from a pre-trained VGG16 model, but I'm having trouble implementing the same solution. Specifically, I want to test the model without training it. I understand that the first dimension corresponds to the batch size, but I'm unsure how to resolve this input shape issue.

Here’s the code I’ve been working with:

# Load pre-trained VGG16 model with ImageNet weights
base_model = VGG16(weights='imagenet', include_top=True)

# Extract features only up to Conv2-2 layer (Layer 5 in VGG16)
output_layer = base_model.get_layer("block2_conv2").output # Conv2-2
feature_extractor = Model(inputs=base_model.input, outputs=output_layer)


feature_extractor.summary()

# input_image.shape = (1, 244, 244, 3)
feature_map = feature_extractor.predict(input_image)

# ValueError: Input 0 of layer "functional_9" is incompatible with the layer: expected shape=(None, 224, 224, 3), found shape=(1, 244, 244, 3)

Continue reading...