FiLo++: Zero-/Few-Shot Anomaly Detection by Fused Fine-Grained Descriptions and Deformable Localization

2025-01-17

AI Summary

Addressing key challenges in zero- and few-shot anomaly detection—including the absence of normal samples from the target class, difficulty in fine-grained localization, and imprecise anomaly descriptions—this paper proposes FusDes-DefLoc, a novel framework comprising two core components. First, FusDes introduces fusion-based fine-grained description generation, enabling class-adaptive semantic characterization of anomalies without relying on target-class normal data. Second, DefLoc employs deformable cross-modal localization via multi-scale deformable cross-modal interaction (MDCI) and position-enhanced patch matching, achieving accurate pixel-level localization even for irregular anomaly regions. By synergistically integrating large language models with Grounding DINO, FusDes-DefLoc performs end-to-end detection and precise localization without requiring any normal samples from the target class. Extensive experiments demonstrate substantial improvements over state-of-the-art methods across multiple benchmarks, significantly enhancing both detection accuracy and localization robustness in zero- and few-shot settings.

Technology Category

• Computer Vision: Multi-modal Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Multimodal Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Anomaly detection methods typically require extensive normal samples from the target class for training, limiting their applicability in scenarios that require rapid adaptation, such as cold start. Zero-shot and few-shot anomaly detection do not require labeled samples from the target class in advance, making them a promising research direction. Existing zero-shot and few-shot approaches often leverage powerful multimodal models to detect and localize anomalies by comparing image-text similarity. However, their handcrafted generic descriptions fail to capture the diverse range of anomalies that may emerge in different objects, and simple patch-level image-text matching often struggles to localize anomalous regions of varying shapes and sizes. To address these issues, this paper proposes the FiLo++ method, which consists of two key components. The first component, Fused Fine-Grained Descriptions (FusDes), utilizes large language models to generate anomaly descriptions for each object category, combines both fixed and learnable prompt templates and applies a runtime prompt filtering method, producing more accurate and task-specific textual descriptions. The second component, Deformable Localization (DefLoc), integrates the vision foundation model Grounding DINO with position-enhanced text descriptions and a Multi-scale Deformable Cross-modal Interaction (MDCI) module, enabling accurate localization of anomalies with various shapes and sizes. In addition, we design a position-enhanced patch matching approach to improve few-shot anomaly detection performance. Experiments on multiple datasets demonstrate that FiLo++ achieves significant performance improvements compared with existing methods. Code will be available at https://github.com/CASIA-IVA-Lab/FiLo.

Problem

• Anomaly Detection
• Zero-shot Learning
• Few-shot Learning

Innovation

• FiLo++
• Grounding DINO
• Zero-shot/Few-shot Anomaly Detection