arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.16451cs.CV
收藏

TinyGLASS: 工业质量控制中的实时自监督传感器异常检测

TinyGLASS: Real-Time Self-Supervised In-Sensor Anomaly Detection

ETH Zurich(苏黎世联邦理工学院);Interdisciplinary Transformation University of Austria(奥地利跨学科转型大学);Swiss Engineering Partners AG(瑞士工程伙伴公司)

Pietro Bonazzi, Rafael Sutter, Luigi Capogrosso, Mischa Buob, Michele Magno

AI总结:

TinyGLASS通过轻量级改进实现实时边缘和传感器异常检测,压缩参数8.6倍,达到94.2%的AUROC并在8MB内存下以20FPS运行。

摘要:

异常检测在工业质量控制中起关键作用,需在标注缺陷样本稀缺的情况下识别缺陷。最近的自监督方法,如GLASS,仅使用无缺陷数据学习正常视觉模式,并在工业基准上表现出色。然而,其计算需求限制了在资源受限的边缘平台上的部署,尤其是在传感器处理架构中。本文介绍TinyGLASS,一种轻量级GLASS框架的改进版本,用于实时边缘和传感器异常检测。所提出架构将原始WideResNet-50主干替换为紧凑的ResNet-18,并引入基于部署的修改,使静态图追踪和INT8量化成为可能。我们评估了该方法在索尼IMX500智能视觉传感器上的性能,利用索尼模型压缩工具包利用传感器处理器。除了在MVTec-AD基准上的性能评估外,我们还研究了训练数据污染的鲁棒性,并引入了名为MMS数据集的定制工业数据集进行跨设备评估。实验结果表明,TinyGLASS在参数压缩8.6倍的同时保持了具有竞争力的检测性能,在MVTec-AD上达到94.2%的图像级AUROC,并在IMX500平台的8MB内存限制下以20FPS运行。系统分析展示了低功耗(每推理4.0mJ)、实时端到端延迟(20FPS)和高能效(470GMAC/J)。此外,模型在中等水平的训练数据污染下表现出稳定的性能。

英文摘要:

Anomaly detection plays a key role in industrial quality control, where defects must be identified despite the scarcity of labeled faulty samples. Recent self-supervised approaches, such as GLASS, learn normal visual patterns using only defect-free data and have shown strong performance on industrial benchmarks. However, their computational requirements limit their deployment on resource-constrained edge platforms, especially within in-sensor processing architectures. This work introduces TinyGLASS, a lightweight adaptation of the GLASS framework designed for real-time edge and in-sensor anomaly detection. The proposed architecture replaces the original WideResNet-50 backbone with a compact ResNet-18 and introduces deployment-based modifications that enable static graph tracing and INT8 quantization. We evaluate the proposed approach on the Sony IMX500 intelligent vision sensor, exploiting the in-sensor processor using the Sony Model Compression Toolkit. In addition to evaluating performance on the MVTec-AD benchmark, we investigate robustness to contaminated training data and introduce a custom industrial dataset, named MMS Dataset, for cross-device evaluation. Experimental results show that TinyGLASS achieves 8.6x parameter compression while maintaining competitive detection performance, reaching 94.2% image-level AUROC on MVTec-AD and operating at 20 FPS within the 8 MB memory constraints of the IMX500 platform. System profiling showcases low power consumption (4.0 mJ per inference), real-time end-to-end throughput (20 FPS), and high energy efficiency (470 GMAC/J). Furthermore, the model demonstrates stable performance under moderate levels of training data contamination.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑