Image-Based Virtual Try-on
Generating a photo of a given person wearing a given garment while keeping the garment's texture, logo and fit, for online fashion retail.
Tier Grades
Tier A: Frontier AI labs
Frontier labs can do this, but the exemplars come from KAIST with OMNIOUS and from Sun Yat-sen University with SenseTime.
Tier B: Indonesian AI labs
IDM-VTON adapts a pretrained diffusion model with a 5-author team, a scale an Indonesian lab can match with local garment data.
Tier C: Institution units such as ITB
CP-VTON and IDM-VTON are university papers with 5 and 6 authors, the usual size of a university unit.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
Released try-on models on a pretrained diffusion base let a delivery team show online retailers a try-on proof of value.
Where this comes from
Anchors
- ComputeTraining compute is not in the sources read.
- DataNot in the sources read; IDM-VTON takes a pair of images, one of the person and one of the garment, and builds on a pretrained diffusion UNet.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 5 authors on IDM-VTON (arXiv 2403.05139) and 6 on CP-VTON (arXiv 1807.07688).
- Time horizonMonths.
- Shelf-life signalsMethods moved from GAN warping (CP-VTON) to diffusion models that keep garment identity (IDM-VTON), General instruction-based image editors with reference images may absorb dedicated try-on models
- Pipeline roleTier C publishes try-on methods on Tier A diffusion bases; Tier B fits them to Indonesian clothing and body types; Tier D runs retail proofs of value.
Where this comes from
Who and What
Links
Papers
- VITON: An Image-based Virtual Try-on NetworkXintong Han et al., arXiv preprint, 2017 founding
- OOTDiffusion: Outfitting Fusion based Latent Diffusion for Controllable Virtual Try-onYuhao Xu et al., arXiv preprint, 2024 state of the art
- Improving Diffusion Models for Authentic Virtual Try-on in the WildYisol Choi et al., ECCV 2024 (per arXiv comment), 2024 state of the art
- CatVTON: Concatenation Is All You Need for Virtual Try-On with Diffusion ModelsZheng Chong et al., ICLR 2025 (per arXiv comment), 2024 state of the art
- TryOnDiffusion: A Tale of Two UNetsLuyang Zhu et al., CVPR 2023 (per arXiv comment), 2023 state of the art
Where this comes from
Sources
- arxiv.org/abs/2403.05139 https://arxiv.org/abs/2403.05139
- arxiv.org/abs/1807.07688 https://arxiv.org/abs/1807.07688
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Image-Based Virtual Try-on", https://aurion.semai.id/topics/image-based-virtual-try-on/, records of October 2026.