Research map / Image generation / Editing and style transfer
Source-Agnostic Image Translation Based on Latent Aware Adaptive Masking
Tomislav Dobrički, Byung-Woo Hong
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 |
|---|---|---|---|
| use dynamic per-step masks instead of constant mask M_0.5T (ablation) | PM-Edit · AFHQ Wild->Dog | LPIPS: 0.309 -> 0.445 (LPIPS), 33.3 -> 17.2 (KID) | Mixed |
| use dynamic per-step masks instead of constant final mask M_0 (ablation) | PM-Edit · AFHQ Wild->Dog | LPIPS: 0.533 -> 0.445 | Improved |
| use time-dependent threshold instead of constant delta_t (ablation) | PM-Edit · AFHQ Wild->Dog | LPIPS: 0.698 -> 0.445 | Improved |
| average discrepancy over N=10 batches instead of N=1 (ablation) | PM-Edit · AFHQ Wild->Dog | LPIPS: 0.617 -> 0.445 | Improved |
| add RePaint inpainting (ablation) | PM-Edit · AFHQ Wild->Dog | KID: 28.0 -> 17.2 | Improved |
| raise threshold coefficient lambda from 1.0 to 1.8 (ablation) | PM-Edit · AFHQ Wild->Dog | FID: FID 54.12 -> 62.89, SSIM 0.453 -> 0.531, LPIPS 0.459 -> 0.406 | Mixed |
| raise lambda from 1.2 to 1.4 (ablation) | PM-Edit · AFHQ Wild->Dog | KID: 17.2 -> 19.4 | Hurt |
| use PM-Edit instead of CycleDiffusion (source-trained) | translation · AFHQ Wild->Dog | KID: 19.5 -> 17.2 | Improved |
| use PM-Edit instead of CycleDiffusion on Cat->Dog | translation · AFHQ Cat->Dog | FID: 58.87 -> 61.31 | Hurt |
| use PM-Edit instead of CycleDiffusion on Male->Female | translation · Celeba-HQ Male->Female | FID: 45.04 -> 50.99 | Hurt |
| use PM-Edit instead of PnPInversion | prompt editing · PIE-Bench | structure distance: 24.29 -> 19.68 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SDEdit — SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations (ICLR 2022) · better on 6 of 6 shared results
- EGSDE — EGSDE: Unpaired Image-to-Image Translation via Energy-Guided Stochastic Differential Equations (NeurIPS 2022) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LMD: Faster Image Reconstruction with Latent Masking Diffusion (AAAI 2024 · similarity 0.87)
- High-Fidelity Diffusion-Based Image Editing (AAAI 2024 · similarity 0.87)
- Simplified and Generalized Masked Diffusion for Discrete Data (NeurIPS 2024 · similarity 0.87)
- Decouple Distortion from Perception: Region Adaptive Diffusion for Extreme-low Bitrate Perception Image Compression (CVPR 2025 · similarity 0.87)
- Masked Autoencoders Are Effective Tokenizers for Diffusion Models (ICML 2025 · similarity 0.87)
- OSCAR: One-Step Diffusion Codec Across Multiple Bit-rates (NeurIPS 2025 · similarity 0.87)
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.
