Research Article

CPGAN :  An Efficient Architecture Designing for Text-to-Image Generative Adversarial Networks Based on Canonical Polyadic Decomposition

Table 2

Comparison between our model and the original model.

ModelOxford-102CUBFlops#Parameters
FIDISFIDIS

Original79.552.66 ± 0.0368.792.88 ± 0.04
Redesigned74.403.68 ± 0.0865.945.03 ± 0.07