Is it possible to apply 2D convolutions to inputs that are not images in Keras?

Question

I would like to apply 2D convolution to a 2D matrix of numbers. This matrix is not an image. However, it seems that for using 2D convolutional layers in Keras we need to specify