A contrastive model fine-tuned on diffusion-generated pairs with matching depth maps and rewritten captions detects structural infringement better than DINOv2, MoCoV3, and SSCD on new SIA and SIR benchmarks.
Diffusion art or digital forgery? investigating data replication in diffu- sion models,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Semantic to Structure: Learning Structural Representations for Infringement Detection
A contrastive model fine-tuned on diffusion-generated pairs with matching depth maps and rewritten captions detects structural infringement better than DINOv2, MoCoV3, and SSCD on new SIA and SIR benchmarks.