Set Features for Fine-grained Anomaly Detection
This work addresses a specific limitation in anomaly detection for scenarios like image and time-series analysis, offering a novel solution that is incremental but impactful for these domains.
The paper tackled the problem of fine-grained anomaly detection where anomalies arise from unusual combinations of normal elements, overcoming limitations of segmentation-based methods by proposing set features that model samples via element distributions, resulting in state-of-the-art performance improvements of +3.4% in image-level logical anomaly detection and +2.4% in sequence-level time-series anomaly detection.
Fine-grained anomaly detection has recently been dominated by segmentation based approaches. These approaches first classify each element of the sample (e.g., image patch) as normal or anomalous and then classify the entire sample as anomalous if it contains anomalous elements. However, such approaches do not extend to scenarios where the anomalies are expressed by an unusual combination of normal elements. In this paper, we overcome this limitation by proposing set features that model each sample by the distribution its elements. We compute the anomaly score of each sample using a simple density estimation method. Our simple-to-implement approach outperforms the state-of-the-art in image-level logical anomaly detection (+3.4%) and sequence-level time-series anomaly detection (+2.4%).