AUM-Art SECAC Lighting Dataset: A Multimodal Collection of AI-Generated Artworks and Prompts for Aesthetic and Lighting Analysis

The AUM-Art SECAC Lighting Dataset presents a multimodal collection of AI-generated artworks and their corresponding text prompts, curated to support research on lighting, aesthetics, and visual\u2013textual embeddings in generative art. Each artwork\u2013prompt pair includes multiple image variants (V1, V2) and associated descriptive prompts reflecting iterative creative refinement. The dataset integrates deep neural embeddings from CLIP and BERT with handcrafted lighting descriptors that quantify global brightness, contrast, and color temperature in LAB space. This combination enables analyses of how textual guidance influences lighting, mood, and composition in AI-generated images.Please cite: N. Ghodke, O. Kursun, Algorithmic Forgetting: AI's Role in Respecting Visual Intellectual Property. Southeastern College Art Conference (SECAC), Session: Generative AI\u2019s Role in Transforming Creative Practices and Cultural Narratives, Cincinnati, Ohio, USA, 22-25 October 2025.