Refine
Document Type
Language
- English (6)
Publication reviewed
- peer review (6)
Keywords
- physically based rendering (2)
- AI-based quality inspection (1)
- ANN inference (1)
- ANN inference, tensorflow lite, embedded systems, benchmarking, response times (1)
- Automatic labeling (1)
- Computer vision (1)
- Physically Based rendering (1)
- Synthetic data generation (1)
- Synthetic images (1)
- Tensorflow Lite (1)
The response time of Artificial Neural Network (ANN)-inference is of utmost importance in embedded applications, particularly continual stream-processing. Predictive maintenance applications require timely predictions of state changes. This study serves to enable the reader to estimate the response time of a given model based on the underlying platform, and emphasizes the relevance of benchmarking generic ANN applications on edge devices. We analyze the influence of net parameters, activation functions as well as single- and multithreading on execution times. Potential side effects such as tact rate variances or other hardware-related influences are being outlined and accounted for. The results underline the complexity of task-partitioning and scheduling strategies while emphasizing the necessity of precise concertation of the parameters to achieve optimal performance on any platform. This study shows that cutting-edge frameworks don’t necessarily perform the required concertations automatically for all configurations, which may negatively impact performance.
In an industrial context, AI-based methods are becoming increasingly important in the optical systems used for identification, inspection and classification. The reasons for this are that AI-based image processing algorithms are easy to use on the operator side and often achieve superior results. E.g. in complex classification tasks. In the sand cast industry, the complexity in optical inspection of cast parts is connected with strong variations in the local surface topography and in the global object geometry change. Despite the great potential of AI-based methods, application is often hindered by the immense effort involved in acquiring a suitable training dataset. This refers not only to the acquisition of the required number of images but also to the tedious labelling. In this work, we investigate the capabilities and limits of synthetic training data on an AI-based optical scanner used to identify and track cast parts. The optical scanner is capable of detecting and classifying a codification specifically designed for the casting industry. By reading the code, the scanner can deduce the specific number of the cast part. For synthetic image generation, we use physically based rendering, which has advantage of full control over all rendering parameters. This allows for both a systematic investigation of the importance of the parameters and, an automatic labelling process of the training datasets. Our results show that, in particular, a detailed geometric modelling of the local surface topography and global object geometry of the pins have a positive influence on the recognition rate of the neural network. With that accuracy rates up to 56 % are achieved using synthetic training datasets, only.
The utilization of uncrewed aerial vehicles (UAVs) in search and rescue (SAR) operations has become increasingly prevalent because the deployment of UAVs is expected to facilitate a higher degree of operational flexibility while simultaneously reducing costs. Currently, commercially available UAVs can be equipped with low-resolution thermal infrared (IR) cameras with typical resolutions of 640 × 512 pixels, which generally are evaluated manually by the SAR teams during an operation. Automatic person detection in IR images still remains a challenge. The objective of the proposed AIResQ dataset is to significantly enhance the performance of object detectors in the IR domain, employed in SAR operations for missing and potentially injured persons. AIResQ comprises 9,788 IR images with a resolution of up to 2048 × 1536 pixels captured from drone perspectives with a handheld camera under
varying weather conditions and in different terrains. Additionally, AIResQ displays persons in atypical poses. In order to test new object detectors in the context of SAR, we established a benchmark dataset stemming from exercises with real drone flights together with SAR organizations.
The response time of Artificial Neural Network (ANN) inference is critical in embedded systems processing sensor data close to the source. This is particularly important in applications such as predictive maintenance, which rely on timely state change predictions. This study enables estimation of model response times based on the underlying platform, highlighting the importance of benchmarking generic ANN applications on edge devices. We analyze the impact of network parameters, activation functions, and single- versus multi-threading on response times. Additionally, potential hardware-related influences, such as clock rate variances, are discussed. The results underline the complexity of task partitioning and scheduling strategies, stressing the need for precise parameter coordination to optimise performance across platforms. This study shows that cutting-edge frameworks do not necessarily perform the required operations automatically for all configurations, which may negatively impact performance. This paper further investigates the influence of network structure on model calibration, quantified using the Expected Calibration Error (ECE), and the limits of potential optimisation opportunities. It also examines the effects of model conversion to Tensorflow Lite (TFLite), highlighting the necessity of considering both performance and calibration when deploying models on embedded systems.
In industrial applications, AI-based methods are gaining increasing significance in optical systems for tasks such as identification, inspection, and classification. The appeal of these methods lies in their operator-friendly implementation and often superior performance, particularly in complex classification tasks. However, despite the substantial potential of AI-based methods, their application is frequently constrained by the significant resources needed to acquire an appropriate training dataset. For instance, in the sand casting industry, the complexity of optical inspection for cast parts is compounded by considerable variation in local surface topography and global object geometry. This challenge pertains not only to the volume of images required but also to the extensive process of accurate labelling. In this work, we investigate the effectiveness and limitations of synthetic training data for an AI-based optical code reader, designed for the identification and tracking of cast parts. This scanner is capable of detecting and classifying a unique code, called Cast Code, developed specifically for the casting industry, enabling the identification of individual cast part numbers. For synthetic image generation, we utilize physically based rendering, which allows comprehensive control over all rendering parameters. This approach facilitates both a systematic investigation of parameter relevance and an automated labelling process for the training datasets. Our results indicate that detailed geometric modelling of both, the local surface topography and the global object geometry of the pins has a notable positive impact on the neural network’s recognition accuracy, achieving accuracy rates of up to 84 %, on real pin images, using synthetic training datasets, only.
Synthetic Data Generation for AI-Based Quality Inspection of Laser Welds in Lithium-Ion Batteries
(2025)
Manufacturing companies are increasingly confronted with critical challenges such as a shortage of skilled labor, rising production costs, and ever-stricter quality requirements. These challenges become particularly acute when defect types exhibit high visual variance, making consistent and accurate inspection difficult. Traditionally, visual inspection of high variance errors is performed manually by human operators—a process that is both costly and prone to errors. Consequently, there is a growing interest in replacing human inspection with AI-based visual quality control systems. However, the adoption of such systems is often hindered by limited access to training data, labor-intensive labeling processes, or the absence of real production data during early development stages. To address these challenges, this paper presents a methodology for training AI models using synthetically generated image data. The synthetic images are created using Physically Based Rendering, which enables precise control over rendering parameters and facilitates automated labeling. This approach allows for a systematic analysis of parameter importance and bypasses the need for large real training datasets. As a case study, the focus is on the inspection of laser welds in battery connectors for fully electric vehicles—a particularly demanding application due to the criticality of each weld. The results demonstrates the effectiveness of synthetic data in training robust AI models, thereby providing a scalable and efficient alternative to traditional data acquisition and labeling methods. The trained binary classifier reaches a precision of 0.94 with a recall of 0.98 solely trained on synthetic data and tested on real image data.
