DMSEC: Diffusion Model Sexually Explicit Content Dataset

The Diffusion Model Sexually Explicit Content Dataset (DMSEC) is a fully synthetic dataset containing sexually explicit images of non-real adults, ensuring that no real individuals are represented and thereby avoiding ethical and privacy concerns. The dataset was generated using four state-of-the-art diffusion models: SDXL 1.0 with the SDXL Refiner 1.0, RealVisXL 5.0, Stable Diffusion 3.5 Medium, and Diffusion 3.5 Large. For each model, 1,080 images were created based on 216 curated prompts and five constant seed values. DMSEC addresses the lack of accessible, ethically safe resources for studying sexually explicit content, enabling systematic assessment of model-specific generation characteristics and supporting the development of robust AI-based content detection systems.