UniFormaly: Towards Task-Agnostic Unified Framework for Visual Anomaly Detection

TL;DR

UniFormaly introduces a unified, task-agnostic framework for visual anomaly detection that leverages novel techniques like Back Patch Masking and top k-ratio feature matching, achieving superior results across diverse datasets.

Contribution

The paper proposes UniFormaly, a universal anomaly detection framework that unifies multiple tasks using innovative methods, reducing resource use and maintenance.

Findings

Achieves state-of-the-art results on various anomaly detection datasets.

Effectively scales to large datasets with reduced memory consumption.

Unifies multiple anomaly detection tasks into a single framework.

Abstract

Visual anomaly detection aims to learn normality from normal images, but existing approaches are fragmented across various tasks: defect detection, semantic anomaly detection, multi-class anomaly detection, and anomaly clustering. This one-task-one-model approach is resource-intensive and incurs high maintenance costs as the number of tasks increases. We present UniFormaly, a universal and powerful anomaly detection framework. We emphasize the necessity of our off-the-shelf approach by pointing out a suboptimal issue in online encoder-based methods. We introduce Back Patch Masking (BPM) and top k-ratio feature matching to achieve unified anomaly detection. BPM eliminates irrelevant background regions using a self-attention map from self-supervised ViTs. This operates in a task-agnostic manner and alleviates memory storage consumption, scaling to tasks with large-scale datasets. Top k-ratio feature matching unifies anomaly levels and tasks by casting anomaly scoring into multiple instance learning. Finally, UniFormaly achieves outstanding results on various tasks and datasets. Codes are available at https://github.com/YoojLee/Uniformaly.