Research map / Image generation / Editing and style transfer
Improving Diffusion Models for Authentic Virtual Try-on in the Wild
Yisol Choi, Sangkyung Kwak, Kyungmin Lee, Hyungwon Choi, Jinwoo Shin
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 |
|---|---|---|---|
| add GarmentNet (parallel UNet feature encoder into self-attention) (ablation) | IDM-VTON · VITON-HD | LPIPS: 0.121 -> 0.102 | Improved |
| add GarmentNet (ablation) | IDM-VTON · VITON-HD | CLIP-I: 0.846 -> 0.883 | Improved |
| use detailed garment captions instead of naive captions | IDM-VTON · VITON-HD (qualitative) | garment consistency | Improved |
| customize with a single person-garment pair | IDM-VTON · In-the-Wild | LPIPS: 0.164 -> 0.150 | Improved |
| customize with a single person-garment pair | StableVITON · In-the-Wild | LPIPS: 0.260 -> 0.259 | No clear effect |
| customize decoder (up-block) layers instead of LoRA (ablation) | IDM-VTON customization · In-the-Wild | CLIP-I: 0.881 -> 0.909 | Improved |
| customize decoder layers instead of all UNet layers (ablation) | IDM-VTON customization · In-the-Wild | CLIP-I: 0.891 -> 0.909 | Improved |
| any customization vs default (SSIM) (ablation) | IDM-VTON · In-the-Wild | CLIP-I: 0.901 -> 0.891 (all UNet) | No clear effect |
| IDM-VTON vs GAN-based GP-VTON on SSIM | IDM-VTON · VITON-HD | SSIM: 0.898 -> 0.870 | Hurt |
| train and test on DressCode, compare IDM-VTON with OOTDiffusion | IDM-VTON · DressCode / VITON-HD cross-dataset | FID: 57.87 -> 12.57 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GP-VTON — GP-VTON: Towards General Purpose Virtual Try-On via Collaborative Local-Flow Global-Parsing Learning (CVPR 2023) · better on 2 of 3 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
- Incorporating Visual Correspondence into Diffusion Model for Virtual Try-On (ICLR 2025 · similarity 0.93)
- Cascaded Diffusion Models for Virtual Try-On: Improving Control and Resolution (AAAI 2025 · similarity 0.93)
- M&M VTO: Multi-Garment Virtual Try-On and Editing (CVPR 2024 · similarity 0.93)
- Texture-Preserving Diffusion Models for High-Fidelity Virtual Try-On (CVPR 2024 · similarity 0.92)
- MV-VTON: Multi-View Virtual Try-On with Diffusion Models (AAAI 2025 · similarity 0.92)
- BooW-VTON: Boosting In-the-Wild Virtual Try-On via Mask-Free Pseudo Data Training (CVPR 2025 · similarity 0.92)
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.
