Intelligent Fault Detection (IFD) has garnered significant attention, with recent advances in AI-empowered predictive maintenance. A key challenge in applying IFD models lies in the interpretability of the methods, since the mechanisms are typically complex and difficult to integrate with data-driven approaches. In addition, the integration of edge devices is an emerging trend, which ensures fault detection and subsequent decision making on the edge, and thus offering an instant response as compared to a conventional centralized server-based architecture. However, to realize Edge-based IFD the primary constraints are low storage capacity and limited computational resources. In this paper, we address various critical challenges in automatic Edge-based IFD for motors in industrial settings, focusing on three key constraints, i.e., (a) limited availability of training data, (b) the lack of method interpretability, and (c) the computational and storage limitations of edge devices. To overcome these challenges, we propose a suite of light weight Physics-Informed (PI) AI algorithms to achieve Edge-based IFD - without compromising detection performance. We validate our proposed methods on experimental data for motor fault detection, and additionally present results from the implementation of these methods on an edge device. We discuss the benefits of our proposed solutions, and give directions for future work. ... Read more

With the exponential growth of video data, individuals, particularly scholars in the fields of history and sociology, are increasingly reliant on video materials. However, the task of locating specific frames within videos remains a laborious and time-consuming endeavor. Advanced machine learning-assisted video processing techniques have emerged, including text-based video searches, video summarization, real-time object detection, and person re-identification. However, distinct from these, the main challenge of retrieving video frames based on given visual content is how to efficiently and accurately pinpoint the instance occurrences. To expedite the process while maintaining retrieval performance, we propose a two-stage approach, combining KeyFrame Extraction (KFE) and Content-based Image Retrieval (CBIR), underpinned a DNN-empowered framework called MoReSo. Our innovations include 1) the integration of improved statistical features with dynamic clustering in the KFE stage and 2) the development of the MoReSo framework, which consists of MobileNet and ResNet backbones with SOA layer to jointly represent video frames, achieving 2.67x increase in efficiency compared to existing solutions. Our framework is evaluated on two datasets: the annotated EHM Historical Database provided by digital history researchers and the widely-used image retrieval benchmark datasets, the Oxford and Paris datasets. The experimental results showcase that the proposed framework and scheme excel among other models in the CBVIR task. We make our code available for further exploration through our GitHub repository. This repository contains the implementation of our model and CBVIR system with a GUI prototype. ... Read more