I tried training it experimentally to see if this was possible.
Since the dataset is limited, the expression of 'boy' is relatively diverse, but the expression of 'girl', especially character representation, is lacking.
