A Novel AI-Based Visual Stimuli Generation Approach for Environment Concept Design

Table 3

The advantages and disadvantages of the four model generation results.

Model	Strengths of model	Weakness of model
PGGAN	The spatial level of the picture is rich; the clarity is high; and the position of the center in the vision basically follows the principle of composition.	The contour line of the visual center is not clear, and its small distinction from the surrounding elements is easily misleading.
WGAN	The pictures have rich spatial levels, high definition, and clear contour lines of the subjects whose position follows the principles of composition so that it is easy to distinguish from other elements.	The distribution of elements in the picture is monotonous, and the diversity is inadequate.
WGAN_gp	The completeness of the pictures is high, and the position of the main objects basically follows the principle of composition.	There is lack of spatial level, small difference between the main object and other elements, as well as lack of clarity in the pictures.
DCGAN	The completeness of the pictures is high with rich spatial levels.	The elements of the picture are ambiguous and illegible, which are arranged in a haphazard manner that does not follow the principles of composition.