arXiv cs.AI· Caterina Gallegati, Monica Bianchini, Franco Scarselli, Vittorio Murino, Barbara Toniella Corradini·· 9 天前AI 评分20
APEX：无假设的基于投影嵌入检验度量，用于生成分布评估
APEX: Assumption-free Projection-based Embedding eXamination Metric for Generative Distribution Evaluation
AI 导读
APEX 是一个基于 Sliced Wasserstein Distance 的无假设生成分布评估框架，用 CLIP 和 DINOv2 作为特征提取器，不依赖传统 FID 等参数化度量。该框架对视觉退化表现出优于现有基线的鲁棒性，并在数据集内和跨数据集上保持稳定，可在域外数据集上实现高稳定性评估。
正文 · 原文
Abstract:As generative models achieve unprecedented visual quality, the gold standard for image evaluation remains traditional feature-distribution metrics (e.g., FID). However, these metrics are provably hindered by the closed-vocabulary bottleneck of outdated features and the assumptive bias of rigid parametric formulations. Recent alternatives exploit modern backbones to solve the feature bottleneck, yet continue to suffer from parametric limitations. To close this gap, we introduce APEX (Assumption-free Projection-based Embedding eXamination), a novel evaluation framework leveraging the Sliced Wasserstein Distance as a mathematically grounded, assumption-free similarity measure. APEX inherits effective scalability to high-dimensional spaces, as we prove with theoretical and empirical evidences. Moreover, APEX is embedding-agnostic and uses two open-vocabulary foundation models, CLIP and DINOv2, as feature extractors. Benchmarking APEX against established baselines reveals superior robustness to visual degradations. Additionally, we show that APEX metrics exhibit intra- and cross-dataset stability, ensuring highly stable evaluations on out-of-domain datasets.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.07786 [cs.CV] |
| (or arXiv:2605.07786v3 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.07786 arXiv-issued DOI via DataCite |
Submission history
From: Barbara Toniella Corradini [view email]
[v1]
Fri, 8 May 2026 14:21:51 UTC (9,371 KB)
[v2]
Mon, 11 May 2026 08:25:22 UTC (16,410 KB)
[v3]
Thu, 8 Oct 2026 15:57:32 UTC (25,178 KB)
来源：arXiv cs.AI · arxiv.org