Research map / Image generation / Autoregressive and latent generation
Reference-Free Quality Assessment for Virtual Try-On via Human Feedback
Yuki Hirakawa, Takashi Wada, Ryotaro Shimizu, Takuya Furusawa, Yuki Saito, Ryosuke Araki, Tianwei Chen, Fan Mo, Yoshimitsu Aoki
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Interleaved Cross-Attention (ICA) (ablation) | VTON-IQA · VTON-QBench test | SRCC: 0.617 -> 0.750 | Improved |
| add Interleaved Cross-Attention (ICA) (ablation) | VTON-IQA · VTON-QBench test | A_micro: 0.747 -> 0.790 | Improved |
| add Interleaved Cross-Attention (ICA) (ablation) | VTON-IQA · VTON-QBench test | R2: 0.372 -> 0.553 | Improved |
| add Interleaved Cross-Attention (ICA) (ablation) | VTON-IQA · VTON-QBench test | PLCC: 0.615 -> 0.751 | Improved |
| fine-tune on VTON-QBench instead of zero-shot DINOv3 | DINOv3 scorer · VTON-QBench test | SRCC: 0.244 -> 0.617 | Improved |
| fine-tune only last half of layers instead of full network (ablation) | VTON-IQA · VTON-QBench test (backbone and fine-tuning ablation) | SRCC: 0.226 -> 0.750 | Improved |
| fine-tune last half instead of last quarter of layers (ablation) | VTON-IQA · VTON-QBench test (backbone and fine-tuning ablation) | SRCC: 0.746 -> 0.750 | No clear effect |
| scale backbone from ViT-B/16 to ViT-L/16 (ablation) | VTON-IQA · VTON-QBench test (backbone and fine-tuning ablation) | SRCC: 0.726 -> 0.750 | Improved |
| add synthetic pairs to real training data (ablation) | VTON-IQA · VTON-QBench test S (train data ablation) | SRCC: 0.642 -> 0.732 | Improved |
| add synthetic pairs to real training data (ablation) | VTON-IQA · VTON-QBench test R (train data ablation) | SRCC: 0.739 -> 0.738 | No clear effect |
| evaluate on unknown VTON models instead of known | VTON-IQA · VTON-QBench cross-group generalization | SRCC: 0.733 -> 0.676 | Hurt |
| evaluate VTON-IQA against human annotator agreement | VTON-IQA · VTON-QBench test (half-annotator labels, 10 runs) | A_micro: 0.791 -> 0.783 | No clear effect |
Closest papers by meaning
- OpenVTON-Bench: A Large-Scale High-Resolution Benchmark for Controllable Virtual Try-On Evaluation (NeurIPS 2026 · similarity 0.90)
- VTON 360: High-Fidelity Virtual Try-On from Any Viewing Direction (CVPR 2025 · similarity 0.89)
- MV-VTON: Multi-View Virtual Try-On with Diffusion Models (AAAI 2025 · similarity 0.88)
- M&M VTO: Multi-Garment Virtual Try-On and Editing (CVPR 2024 · similarity 0.88)
- IMAGDressing-v1: Customizable Virtual Dressing (AAAI 2025 · similarity 0.87)
- BooW-VTON: Boosting In-the-Wild Virtual Try-On via Mask-Free Pseudo Data Training (CVPR 2025 · similarity 0.86)
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.
