Alert-CLIP：面向视频异常检测的异常感知潜在增强 CLIP 表征微调
Alert-CLIP: Abnormality-aware Latent-Enhanced Representation Tuning of CLIP for Video Anomaly Detection
针对 CLIP 文本嵌入空间中正常/异常描述纠缠的问题，提出视频-标签、区域-文本、区域-语义三级对齐微调策略，可直接迁移到任何需要细粒度异常感知的 CLIP 下游任务（如工业缺陷、医学异常筛查）。
原文依据 normal and abnormal descriptions are highly entangled in the text embedding space
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论随着 CLIP 等预训练视觉-语言模型的兴起，通过跨模态推理进行视频异常检测（VAD）已成为一种新兴趋势。然而，我们观察到 CLIP 仍存在异常感知能力薄弱的问题：正常与异常描述在文本嵌入空间中高度纠缠，导致视频特征对两类提示给出几乎无法区分的相似度分数。为解决该问题，我们提出 Alert-CLIP，一个为 VAD 量身定制的异常感知潜在增强微调框架。Alert-CLIP 引入多级对齐策略：（1）视频-标签对齐，重塑语义空间，为异常感知建立粗粒度基础；（2）区域-文本对齐，显式地将异常相关区域与其详细描述关联，以增强细粒度感知；（3）区域-语义对齐，进一步将异常区域与多个难负样本进行对比，增强异常感知判别能力。在四个基准上的大量实验表明，Alert-CLIP 在监督、零样本和开放词表设定下均持续优于原始 CLIP，为未来基于 CLIP 的 VAD 研究提供了坚实基础。
英文原文摘要 1279 字符
With the rise of pre-trained vision-language models such as CLIP, performing video anomaly detection (VAD) through cross-modal reasoning has become an emerging trend. However, we observe that CLIP still suffers from weak abnormality awareness: normal and abnormal descriptions are highly entangled in the text embedding space, causing video features to assign nearly indistinguishable similarity scores to both types of prompts. To address this issue, we propose Alert-CLIP, an abnormality-aware latent-enhanced tuning framework that tailors CLIP for VAD. Alert-CLIP introduces a multi-level alignment strategy: (1) video-label alignment, which reshapes the semantic space to establish a coarse-level foundation for abnormality awareness; (2) region-text alignment, which explicitly associates anomaly-related regions with their detailed descriptions to strengthen fine-grained perception; and (3) region-semantic alignment, which further contrasts anomalous regions against multiple hard negative samples, enhancing abnormality-aware discrimination.Extensive experiments on four benchmarks demonstrate that Alert-CLIP consistently surpasses vanilla CLIP across supervised, zero-shot, and open-vocabulary settings, providing a solid foundation for future CLIP-based VAD research.