![[2602.23013] SubspaceAD: Training-Free Few-Shot Anomaly Detection via Subspace Modeling](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.23013] SubspaceAD: Training-Free Few-Shot Anomaly Detection via Subspace Modeling
Summary
The paper introduces SubspaceAD, a training-free method for few-shot anomaly detection that utilizes subspace modeling to achieve state-of-the-art results without complex training processes.
Why It Matters
SubspaceAD addresses the challenges of few-shot anomaly detection in industrial settings, offering a simpler, efficient alternative to existing methods that often require extensive training and resources. Its high performance on benchmark datasets highlights its potential for practical applications in quality control and automated inspection.
Key Takeaways
- SubspaceAD operates in two stages: feature extraction and PCA modeling.
- Achieves state-of-the-art performance in one-shot and few-shot anomaly detection.
- Eliminates the need for training, prompt tuning, or memory banks, simplifying the process.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.23013 (cs) [Submitted on 26 Feb 2026] Title:SubspaceAD: Training-Free Few-Shot Anomaly Detection via Subspace Modeling Authors:Camile Lendering, Erkut Akdag, Egor Bondarev View a PDF of the paper titled SubspaceAD: Training-Free Few-Shot Anomaly Detection via Subspace Modeling, by Camile Lendering and 2 other authors View PDF HTML (experimental) Abstract:Detecting visual anomalies in industrial inspection often requires training with only a few normal images per category. Recent few-shot methods achieve strong results employing foundation-model features, but typically rely on memory banks, auxiliary datasets, or multi-modal tuning of vision-language models. We therefore question whether such complexity is necessary given the feature representations of vision foundation models. To answer this question, we introduce SubspaceAD, a training-free method, that operates in two simple stages. First, patch-level features are extracted from a small set of normal images by a frozen DINOv2 backbone. Second, a Principal Component Analysis (PCA) model is fit to these features to estimate the low-dimensional subspace of normal variations. At inference, anomalies are detected via the reconstruction residual with respect to this subspace, producing interpretable and statistically grounded anomaly scores. Despite its simplicity, SubspaceAD achieves state-of-the-art performance across one-shot and few-shot settings without...
