The direct integration of robust sensors onto three-dimensional (3D) metallic components is critical for predictive maintenance in industrial systems operating under severe environmental conditions. In this work, we demonstrate a hybrid manufacturing approach combining atmospheric plasma spraying (APS) and femtosecond laser micromachining to fabricate high-definition NiCr/Al2O3/NiCr resistive strain sensors on metallic substrates. Resistive sensors were first produced on flat Al and Al2024 substrates (2D) and subsequently on a stainless steel tube (3D). For sensors deposited on Al2024, uniaxial tensile tests yielded a reliable gauge factor (GF) of 1.51, with very low observable hysteresis and repeatable responses under both stepwise and cyclic loading conditions. The optimized fabrication process was then successfully transferred to the 3D tubular substrate. Three-point bending tests demonstrated reversible, hysteresis-free resistance–strain tracking, confirming pattern integrity and homogeneous coating performance on curved geometries. In addition, the same meander structure functioned as a resistance temperature detector (RTD), exhibiting a highly linear response up to 500 °Cwith temperature coefficients of resistance (TCR) in the range of 1.00 × 10−4 °C−1, thereby demonstrating stable high-temperature operation. Beyond sensitivity values, the key contribution of this work lies in the capability to manufacture stable sensing architectures directly onto arbitrary 3D metallic surfaces, even after exposure to elevated temperatures, for applications in aerospace, automotive, and civil infrastructure components. This paper presents, for the first time, the integration of thermal spray processes with femtosecond laser ablation techniques to fabricate embedded resistive strain with high-definition sensing NiCr patterns on versatile 3D surfaces from object, even after exposure to elevated temperatures. Unlike conventional fabrication approaches, this method enables selective, in situ sensor layer deposition on large-scale components under normal atmospheric conditions without requiring additional post-processing treatments. The performance of the embedded strain sensor was evaluated under deforming forces applied to a 3D tubular structure, demonstrating its structural conformity and sensing capability.
La co-intégration de guides d’onde plasmonique et photonique sur une puce silicium suivant une configuration interférométrique à branche unique, dite bimodale permet d’atteindre une sensibilité inégalée lors de la détection de traces de molécules. Des puces fabriquées à bas coût et un prototype d’instrument complet démontrent leur capacité à détecter des contaminants d’intérêt pour le secteur agricole.
Computer vision tasks such as object detection and segmentation rely on the availability of extensive, accurately annotated datasets. In this work, We present CIA, a modular pipeline, for (1) generating synthetic images for dataset augmentation using Stable Diffusion, (2) filtering out low quality samples using defined quality metrics, (3) forcing the existence of specific patterns in generated images using accurate prompting and ControlNet. In order to show how CIA can be used to search for an optimal augmentation pipeline of training data, we study human object detection in a data constrained scenario, using YOLOv8n on COCO and Flickr30k datasets. We have recorded significant improvement using CIA-generated images, approaching the performances obtained when doubling the amount of real images in the dataset. Our findings suggest that our modular framework can significantly enhance object detection systems, and make it possible for future research to be done on data-constrained scenarios. The framework is available at: github.com/multitel-ai/CIA.
Real-world segmentation tasks in digital pathology require a great effort from human experts to accurately annotate a sufficiently high number of images. Hence, there is a huge interest in methods that can make use of non-annotated samples, to alleviate the burden on the annotators. In this work, we evaluate two classes of such methods, semi-supervised and active learning, and their combination on a version of the GlaS dataset for gland segmentation in colorectal cancer tissue with missing annotations. Our results show that semi-supervised learning benefits from the combination with active learning and outperforms fully supervised learning on a dataset with missing annotations. However, an active learning procedure alone with a simple selection strategy obtains results of comparable quality.