FreqPrompt-AD: Frequency-guided local semantic prompting for zero-shot industrial anomaly detection and segmentation

Abstract

Zero-shot industrial anomaly detection aims to identify and segment defects in unseen categories without target-specific model fitting. CLIP-style vision-language models (VLMs) provide useful semantic transferability, but their global image-text alignment often emphasizes object-level semantics and weakens the response to local high-frequency defects. To address this granularity mismatch, we propose FreqPrompt-AD, a training-free framework that constructs local-frequency semantic evidence with a frozen VLM. Frequency-aware prompt interaction (FPI) converts high-frequency residuals into semantic-gated spatial guidance for patch-text anomaly evidence; local semantic calibration (LSC) removes object-level semantic dominance and builds image-adaptive normality prototypes; and text-anchored decision (TAD) integrates semantic and prototype-deviation evidence into image-level anomaly scores and dense anomaly maps. We further evaluate FreqPrompt-AD-FT, an adapter-enhanced variant in which only a lightweight adapter is optimized while the VLM backbone and text encoder remain frozen. Full-scale experiments on MVTec AD, VisA, BTAD, and MPDD show that FreqPrompt-AD achieves the strongest average performance among the compared zero-shot VLM-based detectors in our local full-scale evaluation, improving average image-level and pixel-level AUROC over DLVP-CLIP by 2.05 and 2.15 percentage points, respectively. Additional frequency-prior comparisons, robustness evaluations, feature interaction analysis, and qualitative results validate the effectiveness and practical applicability of the proposed local-frequency semantic evidence formulation.