The recent rapid development of deep learning has laid a milestone in visual anomaly detection (VAD). In this paper, we provide a comprehensive review of deep learning-based visual anomaly detection techniques, from the perspectives of neural network architectures, levels of supervision, loss functions, metrics and datasets. In addition, we extract the new setting from industrial manufacturing and review the current VAD approaches under our proposed our new setting. Moreover, we highlight several opening challenges for visual anomaly detection. The merits and downsides of representative network architectures under varying supervision are discussed. Finally, we summarize the research findings and point out future research directions. More resources are available at https://github.com/M-3LAB/awesome-industrial-anomaly-detection

翻译：近年来深度学习的迅猛发展为视觉异常检测领域树立了里程碑。本文从神经网络架构、监督程度、损失函数、评估指标与数据集等多个维度，对基于深度学习的视觉异常检测技术进行了全面综述。此外，我们提炼了工业制造中的新场景，并在所提出的新框架下回顾了当前的视觉异常检测方法。同时，我们着重指出了视觉异常检测面临的若干开放性挑战，并对不同监督程度下代表性网络架构的优缺点进行了探讨。最后，我们总结了研究发现并展望了未来研究方向。更多资源请访问 https://github.com/M-3LAB/awesome-industrial-anomaly-detection