OpenVTON-Bench：面向可控虚拟试穿评估的大规模高分辨率基准

View PDF HTML (experimental)

Abstract:Recent advances in diffusion models have significantly elevated the visual fidelity of Virtual Try-On (VTON) systems, yet reliable evaluation remains a persistent bottleneck. Traditional metrics struggle to quantify fine-grained texture details and semantic consistency, while existing datasets fail to meet commercial standards in scale and diversity. We present OpenVTON-Bench, a large-scale benchmark comprising approximately 100K high-resolution image pairs (up to $1536\times1536$). The dataset combines Gemini-powered dense captioning with stratified sampling across 20 visual clusters obtained by DINOv3 K-Means. To support reliable evaluation, we propose a multi-modal protocol that measures VTON quality along five interpretable dimensions: background consistency, identity fidelity, texture fidelity, shape plausibility, and overall realism. The protocol integrates VLM-based semantic reasoning with a novel Multi-Scale Representation Metric based on SAM3 segmentation and morphological erosion, enabling the separation of boundary alignment errors from internal texture artifacts. Under the specified evaluation rubric, our representation metric shows strong agreement with human model rankings (Kendall's $\tau$ of 0.833 vs. 0.611 for SSIM). Code and dataset are available at this https URL.

Comments:	Accepted at NeurIPS 2026, Evaluations and Datasets Track
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2601.22725 [cs.CV]
	(or arXiv:2601.22725v5 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2601.22725 arXiv-issued DOI via DataCite

Submission history

From: Jin Li [view email]
[v1] Fri, 30 Jan 2026 08:58:00 UTC (39,431 KB)
[v2] Mon, 23 Mar 2026 05:00:18 UTC (39,432 KB)
[v3] Wed, 6 May 2026 05:56:20 UTC (24,653 KB)
[v4] Wed, 10 Jun 2026 07:22:12 UTC (24,653 KB)
[v5] Sat, 10 Oct 2026 04:04:40 UTC (27,481 KB)