Pre-trained mode layer, the description of the “Input Shape” parameter says
Optional shape tuple, only to be specified if
include_top is false (otherwise the input shape has to be
(224, 224, 3) (with
tf dim ordering) or
(3, 224, 244) (with
th dim ordering). it should have exactly 3 inputs channels, and width and height should be no smaller than 197. e.g.
(200, 200, 3) would be one valid value.
include_top is false, not specifying the input shape does not stop the Model to be valid and it can be trained. It seem like DLS automatically fills that parameter.