Language-Guided Open-World Anomaly Segmentation

TL;DR

Clipomaly introduces a zero-shot, CLIP-based method for open-world anomaly segmentation in autonomous driving, enabling detection and naming of unknown objects without additional training.

Contribution

It is the first CLIP-based approach that dynamically extends vocabulary at inference time for anomaly segmentation, surpassing existing methods in performance.

Findings

Achieves state-of-the-art results on anomaly segmentation benchmarks.

Provides human-interpretable labels for unknown objects.

Operates without anomaly-specific training data.

Abstract

Open-world and anomaly segmentation methods seek to enable autonomous driving systems to detect and segment both known and unknown objects in real-world scenes. However, existing methods do not assign semantically meaningful labels to unknown regions, and distinguishing and learning representations for unknown classes remains difficult. While open-vocabulary segmentation methods show promise in generalizing to novel classes, they require a fixed inference vocabulary and thus cannot be directly applied to anomaly segmentation where unknown classes are unconstrained. We propose Clipomaly, the first CLIP-based open-world and anomaly segmentation method for autonomous driving. Our zero-shot approach requires no anomaly-specific training data and leverages CLIP's shared image-text embedding space to both segment unknown objects and assign human-interpretable names to them. Unlike open-vocabulary methods, our model dynamically extends its vocabulary at inference time without retraining, enabling robust detection and naming of anomalies beyond common class definitions such as those in Cityscapes. Clipomaly achieves state-of-the-art performance on established anomaly segmentation benchmarks while providing interpretability and flexibility essential for practical deployment.