LogicalDefender: Discovering, Extracting, and Utilizing Common-Sense Knowledge

arXiv:2403.11570v1 Announce Type: new
Abstract: Large text-to-image models have achieved astonishing performance in synthesizing diverse and high-quality images guided by texts. With detail-oriented conditioning control, even finer-grained spatial control can be achieved. However, some generated images still appear unreasonable, even with plentiful object features and a harmonious style. In this paper, we delve into the underlying causes and find that deep-level logical information, serving as common-sense knowledge, plays a significant role in understanding and processing images. Nonetheless, almost all models …

abstract arxiv control cs.cv diverse features generated however image images knowledge object paper performance quality sense spatial style text text-to-image type