CURVTON‑205K: A Mask‑Free Difficulty‑Stratified Synthetic Dataset for Robust and Scalable Virtual Try‑On

WACV 2027
Dipan Bartaula1,5,† Anuj Shrestha1 Ankit Belbase1 Anish Bishwakarma1 Saroj Poudel1 Samip Ghimire1 Shashank Bhatta1 Eric Arazo2 Binod Bhattarai2,3,4
1Department of Electronics and Computer Engineering, IOE Pulchowk Campus, Tribhuvan University, Lalitpur, Nepal 2FogSphere, London, UK 3University College London, London, UK 4School of Natural and Computing Sciences, University of Aberdeen, Aberdeen, UK 5NAAMII, Kathmandu, Nepal
†Corresponding author: deepanzcreed001@gmail.com
Tribhuvan University FogSphere University College London University of Aberdeen NAAMII
Paper Code Dataset arXiv (coming soon)
Source person
Source personIp
In-shop garment
In-shop garmentIc
Synthesized try-on
VLM-accepted
Synthesized try-onIt
Paper Fig. 1
CURVTON-205K triplets: a source person, an in-shop garment and the synthesized try-on, all at 1024×1024 and generated without cloth segmentation masks or dense pose (first example: paper Fig. 1; the others come from all three difficulty tiers).
205K
curated training triplets
4.5K
held-out test triplets, human-verified
200
garment types incl. traditional & long-tail
3
difficulty tiers: easy, medium, hard

Abstract

Recent advances in 2D virtual try-on (VTON) have largely focused on architectural innovation, yet robust performance in diverse real-world conditions remains challenging. We argue that failure modes in VTON, such as inaccurate occlusion handling, unconstrained poses, and poor coverage of under-represented garments and appearance factors, are tightly coupled to dataset design choices and long-tail data regimes. We present a scalable data generation pipeline that coordinates specialized editing and generation models with vision-language model (VLM) guidance to synthesize triplets without requiring parsing masks or dense-pose annotations.

Using this pipeline, we create CURVTON-205K, a curated synthetic dataset of 205K triplets: source person, target garment, and corresponding try-on image. CURVTON-205K is explicitly designed to broaden coverage across garments, person appearance, poses, occlusions, and visual conditions. Additionally, CURVTON-205K is organized into three difficulty tiers, enabling controlled evaluation under varying levels of difficulty and supporting difficulty-aware experimentation. We analyze the dataset diversity and demonstrate its potential to improve try-on fidelity across four VTON architectures and its cross-dataset generalization.

Method

Garments, people and difficulty edits are generated in separate streams, merged into try-on triplets, and checked by a VLM that sends failures back for regeneration.

CURVTON-205K data generation pipeline (paper Fig. 2)
1Cloth stream
Data generation pipeline (paper Fig. 2). Clothing, person and editing streams produce garments, source persons and difficulty-tiered edits; the merging stream synthesizes the try-on with FLUX.2 [klein] 9B and a Qwen3-VL 32B assessor accepts it or returns an improved prompt for another round (up to 4).

The feedback loop, attempt by attempt

Four successive generations per example, taken from the paper (Figs. 13 & 15). The VLM assessor checks every attempt; a rejected try-on goes back to the generator with a refined prompt, and the last attempt is the one it accepts.

Source personPerson
Target garmentGarment
FLUX.2 [klein] 9Btry-on generatorBase VTON prompt
Generated try-on attemptAttempt 1Waiting for the generator…
Qwen3-VL 32BVLM assessor
  • Garment faithfulness
  • Placement plausibility
  • Semantic consistency
  • No visible artifacts
Attempts

Pick an example above; each one plays through its attempts.

Across the dataset, the loop removes most artifacts

Artifact rate per refinement round

% of try-on candidates with artifacts, VLM acceptance threshold 0.8 (Table 6)

Three difficulty tiers

Each source person is edited with tier-specific instructions that raise pose deviation, occlusion, clutter and viewpoint change, giving 100K edited people per tier. The tiers show where a model breaks, not just its average score.

Comparison with existing datasets

CURVTON-205K is the only dataset here that is mask-free, in-the-wild, paired, difficulty-tiered, and diverse in both garments and people (Table 1).

Diversity coverage: Limited = neither garment nor person-appearance diversity targeted; Moderate = one; Broad = both. All listed datasets are public except WildVTON.

Results

Training on CURVTON-205K improves models on its own test set, transfers to real benchmarks, and wins in-the-wild comparisons.

Zero-shot (pretrained)Trained on CURVTON-205K

CURVTON-205K test set (Tables 2 & 8). UB/LB = upper/lower body, G = full-body garments, Tr/NTr = traditional/non-traditional. Hard stays the hardest tier after training.

Real baselineCURVTON-205K onlyReal + CURVTON

Train on real data, CURVTON-205K, or both; test protocols fixed (Table 3).

OOTDiffusion trained on CURVTON-205K is preferred by human raters for both try-on quality (35%) and realism (28%), and scores highest on CLIP-I and the VLM evaluator.

OOTDiffusion trained on each data source; pretrained FLUX and IDM-VTON as references (Table 5). Human preference: blind pairwise votes from 20 annotators, normalized to 100% per criterion.

Explore the data

Compare any source person with its try-on, then browse more try-ons and the garments behind them. All samples come from the public sample split on Hugging Face.

Source person
Synthesized try-on
Source personTry-on Drag to compare
Target garmentEasy
 
1 / 8

Drag the handle (or use ← →) to compare the source person with the synthesized try-on; the strip picks examples from all three difficulty tiers.

Hover a card to see the source person; clickTap a card to open the side-by-side comparison.

Hover to peek at the try-on, clickTap to compare. The garment dictionary covers 200 types, from common Western clothing to long-tail and traditional garments. Each garment here opens the try-on it was used for.

Dataset

Released on Hugging Face as six tar archives (difficulty × presentation), about 108 GB in total, plus a small sample split.

5,627
downloads, all time
326
downloads, last 30 days
209.5K
triplets (205K train + 4.5K test)
~108 GB
6 archives, JPEG triplets
Hugging Face stats as of Oct 3, 2026

Download

from huggingface_hub import snapshot_download

# sample split only (remove allow_patterns for all)
path = snapshot_download(
    repo_id="curvton2/curvton2",
    repo_type="dataset",
    allow_patterns="sample_dataset/*",
)

Layout

easy/ medium/ hard/
└── female/ male/
    ├── cloth_image/           # garment  I_c
    ├── initial_person_image/  # person   I_p
    └── tryon_image/           # try-on   I_t

Matching file names pair a triplet. Code for the pipeline, tier dictionaries, evaluation and diversity metrics: github.com/curvton/CURVTON-205K_CODE.

Citation & license

BibTeX

@inproceedings{bartaula2027curvton205k,
  title  = {{CURVTON-205K}: A Mask-Free
            Difficulty-Stratified Synthetic
            Dataset for Robust and Scalable
            Virtual Try-On},
  author = {Bartaula, Dipan and Shrestha, Anuj
            and Belbase, Ankit and Bishwakarma,
            Anish and Poudel, Saroj and Ghimire,
            Samip and Bhatta, Shashank and
            Arazo, Eric and Bhattarai, Binod},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference
               on Applications of Computer Vision (WACV)},
  year   = {2027}
}

License

CC BY-NC 4.0

The dataset is released under Creative Commons Attribution-NonCommercial 4.0, with the additional non-commercial restrictions of the FLUX.2 [klein] 9B model used to generate it.

Intended for academic, non-commercial research: virtual try-on, garment transfer, controllable image generation, curriculum learning and benchmarking.

Not for:

  • identity, face or biometric recognition, or sensitive-attribute inference;
  • surveillance, medical or clinical decisions;
  • deceptive or non-consensual edited imagery.

Female/male labels organize the data only; they are not identity or demographic annotations.