Alexandre Durupt

The field of industrial defect detection using machine learning and deep learning is a subject of active research. Datasets, also called benchmarks, are used to compare and assess research results. There is a number of datasets in industrial visual inspection, of varying quality. Thus, it is a difficult task to determine which dataset to use. Generally speaking, datasets which include a testing set, with precise labeling and made in real-world conditions should be preferred. We propose a study of existing benchmarks to compare and expose their characteristics and their use-cases. A study of industrial metrics requirements, as well as testing procedures, will be presented and applied to the studied benchmarks. We discuss our findings by examining the current state of benchmarks for industrial visual inspection, and by exposing guidelines on the usage of benchmarks.

The fourth industrial revolution is shaping a new industrial landscape. A variety of technologies related to software, information and communication technologies embody a ubiquitous digital and connectivity era. These technologies enable the creation of new products with the integration of connectivity, data collection and processing capacities which require combining engineering disciplines. Increasing product multidisciplinarity compels companies to adapt their product development practices. The scientific literature offers a variety of concepts and techniques to support multidisciplinary product development. This paper seeks to organize the landscape of concepts and techniques available for multidisciplinary product development. An extensive literature review was conducted, and 236 concepts and techniques were identified. Multidisciplinary products of interest deal with both software and hardware development and can be encountered through the denominations of cyber-physical systems, mechatronics and smart products and systems. An in-depth analysis led to the classification and mapping, for each product denomination, of the concepts and techniques available to support their development. The classification relies on a four-level model paired with a decision tree to thoroughly sort the variety of concepts and techniques into the approach, process, method, and tool levels. The mapping between the sorted concepts and techniques enabled the generation of graphical representations called cartographies. These cartographies serve to support companies’ transformation towards the fourth industrial revolution from the product development perspective by giving a general overview of the related literature, and guiding them in the identification of the most suitable approaches, processes, methods and tools.

L'évolution des techniques de fouilles de données, ainsi que l'accroissement des capacités de stockage et de calcul suscite, dans tous les domaines, un intérêt pour les données produites. En ce sens, le (...)