TokenCLIP: Token-wise Prompt Learning for Zero-shot Anomaly Detection

TL;DR

TokenCLIP introduces a token-wise prompt learning framework that dynamically aligns visual tokens with specialized textual subspaces using optimal transport, significantly improving zero-shot anomaly detection across diverse objects and domains.

Contribution

It proposes a novel token-wise adaptation method with dynamic alignment via optimal transport, enabling fine-grained, efficient zero-shot anomaly detection.

Findings

Outperforms existing zero-shot anomaly detection methods.

Effectively captures varied anomaly semantics across different objects.

Demonstrates superior performance on multiple benchmark datasets.

Abstract

Adapting CLIP for anomaly detection on unseen objects has shown strong potential in a zero-shot manner. However, existing methods typically rely on a single textual space to align with visual semantics across diverse objects and domains. The indiscriminate alignment hinders the model from accurately capturing varied anomaly semantics. We propose TokenCLIP, a token-wise adaptation framework that enables dynamic alignment between visual and learnable textual spaces for fine-grained anomaly learning. Rather than mapping all visual tokens to a single, token-agnostic textual space, TokenCLIP aligns each token with a customized textual subspace that represents its visual characteristics. Explicitly assigning a unique learnable textual space to each token is computationally intractable and prone to insufficient optimization. We instead expand the token-agnostic textual space into a set of orthogonal subspaces, and then dynamically assign each token to a subspace combination guided by semantic affinity, which jointly supports customized and efficient token-wise adaptation. To this end, we formulate dynamic alignment as an optimal transport problem, where all visual tokens in an image are transported to textual subspaces based on semantic similarity. The transport constraints of OT ensure sufficient optimization across subspaces and encourage them to focus on different semantics. Solving the problem yields a transport plan that adaptively assigns each token to semantically relevant subspaces. A top-k masking is then applied to sparsify the plan and specialize subspaces for distinct visual regions. Extensive experiments demonstrate the superiority of TokenCLIP.