Image data is numeric data.
There are, however, times when you want to change the numeric representation, such as by using im2double()
Some layers permit rescaling; see for example https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.sequenceinputlayer.html
Are you possibly asking about having it interpret a graph? If so then see the File Exchange for the tag "digitize" -- I tagged a number of programs last year.
Possibly OCR (Optical Character Recognition) could also be important for interpreting graphs. That has more often been done with shallow networks instead of deep networks, but deep networks could also be used, especially for hand-written material. Deep networks could also be important for interpreting blueprints and mechanical drawings.
