Yasser Abd Djawad, Dary Mochamad Rifqie, Ridwansyah, Supriadi, Saharuddin Ronge Sokku, Andi Baso Kaswar, Ma’Ruf Idris
Objective: The increasing use of artificial intelligence (AI) in various fields has increased the need for a large amount of data. A device with adequate computational power is required to manage the data and produce an output with high processing speed and satisfactory accuracy. Moreover, the use of several embedded-system devices for neural networks (NNs) is constrained by low processor and memory capacity. Several embedded-system devices with improved processor capabilities have been developed for NN data processing. Materials and method: In this study, the capabilities of an embedded-system device for NNs in health applications was analyzed; namely, the detection of X-ray images of patients with pneumonia using a convolutional neural network (CNN) was tested. Two-dimensional CNN architectures with various parameters, including color depth, layers, filters, kernels, and quantization, were employed. The outcome was expressed in terms of accuracy, inference time, RAM, and flash consumption. Results and discussion: The results revealed a significant positive association between all output metrics and the number of filters. However, in some situations, the RAM and flash utilization of the embedded system exceeded its capacity, making it unusable. Thisfinding indicates that the inference time and memory are influenced by the number of layers, filters and quantization. Conclusion:Thus, the use of embedded-system devices for CNN can be done with proper hyperparameter tuning. © 2025, Pontificia Universidad Javeriana. All rights reserved.
Universitas Negeri Makassar, Indonesia; Electronics Department, Universitas Negeri Makassar, Indonesia
