InDecGAN: Learning to Generate Complex Images From Captions via Independent Object-Level Decomposition and Enhancement | IEEE Journals & Magazine | IEEE Xplore