Answers are grounded in this paper and cite their sources.
To address the issues of large scenes and detail attributions for generating remote sensing images (RSIs), this study proposes a multiple conditions-guided diffusion (MCGD) model. First, a prior object image control (POIC) module based on multipoint optimization, which models common ground objects, is proposed as a condition of the diffusion model. Then, a caption condition encoding and control (CCEC) module, which mines the caption semantics of the generated image, is designed to construct the semantic space and realize features' cross-modal transformation from text to image. Finally, this study proposes a visual context perception module based on attention, which deeply integrates the conditional features of POIC and CCEC to enhance the fine-grained RSI generation. Experiments show that MCGD can obtain CLIP-scores of 25.23 and 31.09, FID values of 44.26 and 10.42, and IS values of 4.6 and 7.8 on RSICD and NWPU-Captions datasets, respectively, which proves its effectiveness.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: