AI Paper Atlas

Research map / Image generation / Editing and style transfer

High-Fidelity Virtual Try-On beyond Paired Data Scarcity via Diffusion-based Cycle-Consistent Learning

Jia Wu, Yijing Dai, Tingfeng Cao, Meiling Wu, Tao Luo, Jian Dong Zhang, Guangming Lu, Xiaoyi Zeng

CVPR 2026 · Image generation › Editing and style transfer · approach: Virtual try-on

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 changedWhereMeasuredEffect
add cycle-consistent learning on in-the-wild images to UDiT (instead of typical pseudo-pair fine-tuning) (ablation)CCVTON · VITON-HD (paired)FID: 6.439 -> 5.148Improved
fine-tune UDiT on MCFO-filtered pseudo-paired in-the-wild data with typical pair-based training (ablation)UDiT · VITON-HD (paired)FID: 5.995 -> 6.439Hurt
add garment-aware mask generation at inference (ablation)CCVTON (Exp.4) · VITON-HD (paired)FID: 5.148 -> 4.814Improved
add garment-aware mask generation at inference (ablation)CCVTON · DressCode (paired)FID: 3.411 -> 2.930Improved
add garment-aware mask generation at inference (ablation)CCVTON · VITON-HD (unpaired)KID: 0.518 -> 0.561Hurt
add UDiT joint try-on and try-off training on paired data (ablation)Flux.1 Fill base · VITON-HD (paired)SSIM: 0.884 -> 0.840Hurt
add GAMG and cloth-bbox mask to base model without UDiT or CCL (ablation)Flux.1 Fill base · VITON-HD (paired)FID: 5.995 -> 5.890Improved
use perceptual VGG regularization on filtered pseudo garments instead of using them as inputsCCL try-off branch · VITON-HDFIDImproved
CCVTON versus prior methods across datasetsVTON methods · DressCode (unpaired)FID: 6.540 -> 5.820Improved

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy