Research map / Image generation / Autoregressive and latent generation
Understanding and Mapping Natural Beauty
Scott Workman, Richard Souvenir, Nathan Jacobs
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train on the distribution of human ratings (multinomial likelihood) instead of the mean rating (ablation) | GoogLeNet scenicness CNN · ScenicOrNot | K-S match rate: 14.8% -> 58.4% | Improved |
| train with cross-entropy to normalized rating distribution instead of mean rating (ablation) | GoogLeNet scenicness CNN · ScenicOrNot | K-S match rate: 14.8% -> 50.0% | Improved |
| use multinomial likelihood instead of normalized-distribution cross-entropy (ablation) | GoogLeNet scenicness CNN · ScenicOrNot | K-S match rate: 50.0% -> 58.4% | Improved |
| train on distributions rather than mean rating (ablation) | GoogLeNet scenicness CNN · ScenicOrNot | nDCG: .9780 -> .9745 | No clear effect |
| use multinomial instead of Distribution loss (ablation) | GoogLeNet scenicness CNN · ScenicOrNot | nDCG: .9678 -> .9745 | No clear effect |
| add co-located overhead imagery (CVH) instead of nearest ground image (1NN) (ablation) | scenicness mapping · CVSoN scenicness > 7 | AUC: 64.38% -> 68.51% | Improved |
| add overhead imagery (CVH) instead of locally weighted average of ground images (ablation) | scenicness mapping · CVSoN scenicness > 7 | AUC: 66.86% -> 68.51% | Improved |
| crop out vehicles to maximize predicted scenicness | Multinomial network · example image | predicted scenicness: 5.0 -> 7.3 | Improved |
Closest papers by meaning
- VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations (ICLR 2026 · similarity 0.82)
- Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping (CVPR 2026 · similarity 0.82)
- Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment (AAAI 2026 · similarity 0.82)
- Aligning Vision Models with Human Aesthetics in Retrieval: Benchmarks and Algorithms (NeurIPS 2024 · similarity 0.82)
- The Photographer's Eye: Teaching Multimodal Large Language Models to See, and Critique Like Photographers (CVPR 2025 · similarity 0.81)
- FiVA: Fine-grained Visual Attribute Dataset for Text-to-Image Diffusion Models (NeurIPS 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
