Research map / Image generation / Autoregressive and latent generation
Enhancing Plausibility Evaluation for Generated Designs with Denoising Autoencoder
Jiajie Fan, Amal Trigui, Thomas Bäck, Hao Wang
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 |
|---|---|---|---|
| replace Inception-V3 encoder in FID with a denoising autoencoder encoder (FDD) | FID · BIKED model ranking (DDPM, DDIM, EDM) | agreement with human plausibility ranking: FID ranks EDM best at 7.84; FDD gives EDM worst at 111.25 | Improved |
| use FDD instead of FID and FD_DINOv2 | generative design evaluation · BIKED, Seeing3DChairs, FFHQ, ImageNet sensitivity test | penalty on structural errors vs visual artifacts | Improved |
| use FDD instead of Topology Distance (TD) | plausibility metric · Seeing3DChairs and FFHQ sensitivity test | stability across noise types | Improved |
| use Frechet distance on DAE features instead of MMD (KDD) or topology distance (TDD) (ablation) | DAE-based metrics · BIKED sensitivity test | stability and structural vs visual discrimination | Improved |
| train the DAE on BIKED instead of ImageNet (ablation) | FDD · BIKED sensitivity test | sensitivity to structural disturbance | No clear effect |
| use DAE-based metrics FDD, KDD, and FDD (BIKED) in model ranking (ablation) | DAE-based metrics · BIKED model ranking | correlation with human evaluation error | Improved |
| FDD under escalating disturbance levels | FDD · BIKED and Seeing3DChairs, 1000 images | consistency with increasing disturbance | Improved |
Closest papers by meaning
- Rethinking FID: Towards a Better Evaluation Metric for Image Generation (CVPR 2024 · similarity 0.89)
- Representation Fréchet Loss for Visual Generation (NeurIPS 2026 · similarity 0.87)
- Fréchet Wavelet Distance: A Domain-Agnostic Metric for Image Generation (ICLR 2025 · similarity 0.87)
- Anomaly Score: Evaluating Generative Models and Individual Generated Images based on Complexity and Vulnerability (CVPR 2024 · similarity 0.85)
- LL-Bench: Rethinking Low-Level Vision Evaluation in the Era of Large-Scale Generative Models (NeurIPS 2026 · similarity 0.85)
- MIND: Monge Inception Distance for Generative Models Evaluation (NeurIPS 2026 · similarity 0.84)
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.
