Accurate and efficient polyp segmentation in colonoscopy images is crucial for early colorectal cancer detection. While traditional methods often rely on one-stage early fusion techniques, this paper explores the adva...
详细信息
Energy consumption forecasting is vital for the development of smart grids. Currently, much of the research in this field concentrates on forecasting energy consumption at the individual household level. Few methods f...
详细信息
In recent years, many fusion algorithms based on multi-scale transform or neural networks have been proposed to improve medical image fusion (MIF) performance. However, there is still enormous potential to explore the...
详细信息
ISBN:
(纸本)9798350344868;9798350344851
In recent years, many fusion algorithms based on multi-scale transform or neural networks have been proposed to improve medical image fusion (MIF) performance. However, there is still enormous potential to explore the combination of different fusion theories. In this paper, we propose a novel MIF framework to integrate powerful feature representation abilities of the deeplearning model and accurate frequency decomposition characteristics of discrete wavelet transform (DWT). Firstly, a multi-scale encoder-decoder network is well-trained to extract feature information in different scales and achieve efficient image reconstruction. In particular, DWT is introduced into each scale to decompose the extracted features into high- and low-frequency sub-bands for information preservation during down-sampling. An elaborate feature fusion process is designed to achieve multi-scale fusion while merging different frequency sub-bands. Experiment results on benchmark datasets demonstrate that the proposed fusion framework outperforms current state-of-the-art methods with comparable time complexity in both objective and subjective evaluation.
The extend of tool wear significantly affects blanking processes and has a decisive impact on product quality and productivity. For this reason, numerous scientists have addressed their research to wear monitoring sys...
详细信息
The extend of tool wear significantly affects blanking processes and has a decisive impact on product quality and productivity. For this reason, numerous scientists have addressed their research to wear monitoring systems in order to identify or even predict critical wear at an early stage. Existing approaches are mainly based on indirect monitoring using time series, which are used to detect critical wear states via thresholds or machine learning models. Nevertheless, differentiation between types of wear phenomena affecting the tool during blanking as well as quantification of worn surfaces is still limited in practice. While time series data provides partial insights into wear occurrence and evolution, direct monitoring techniques utilizing image data offer a more comprehensive perspective and increased robustness when dealing with varying process parameters. However, acquiring and processing this data in real-time is challenging. In particular, high dynamics combined with increasing strokes rates as well as the high dimensionality of image data have so far prevented the development of direct image-based monitoring systems. For this reason, this paper demonstrates how high -resolution images of tools at 600 spm can be captured and subsequently processed using semantic segmentation deeplearning algorithms, more precisely Fully Convolutional Networks (FCN). 125,000 images of the tool are taken from successive strokes. Selected images are labeled pixel by pixel according to their wear condition and used to train a FCN (U-Net). The approach presented offers the possibility of spatially monitoring wear in high-speed blanking operations in real-time.
Currently, tumours rank as the second most common cancer kind. A great number of people are at risk because of cancer. In order to diagnose tumours like brain tumours, the medical sector requires a method that is quic...
详细信息
This review paper delves into the rapidly evolving field of driver emotion detection, with a specific focus on the contributions of deeplearning methodologies and the diverse datasets that facilitate this research. F...
详细信息
Cardiac arrest has been a leading cause of mortality worldwide, with limited opportunities for intervention. This project introduces a novel machine-learning approach to predict and process cardiac arrest risk in high...
详细信息
The use of IoT systems in industrial environments provides tremendous benefits and economic value leading to an exponential rise in their adoption. Their extended use, however, does not come without concerns related t...
详细信息
ISBN:
(纸本)9783031783791;9783031783807
The use of IoT systems in industrial environments provides tremendous benefits and economic value leading to an exponential rise in their adoption. Their extended use, however, does not come without concerns related to potential security threats, thereby creating an obstacle in their further use in the field. To address these security concerns, we introduce a specialized Industrial Intrusion Detection System (I2DS). Our proposed system merges the capabilities of deeplearning (DL) with FPGA-based hardware acceleration techniques, enabling it to detect subtle anomalies and potential cyber threats that may evade conventional rule-based intrusion detection systems (IDS) in an effective way. More specifically, by implementing the system on FPGA hardware, we achieve low-latency, high-throughput processing of network traffic, essential for real-time intrusion detection in industrial settings. Our architecture is scalable and can be adapted according to network bandwidth requirements, while remaining lightweight, making it an ideal solution for the stringent resource constraints often encountered in IoT environments. The proposed solution has been validated with the modbus TON-IoT dataset, achieving up to two orders of magnitude higher performance compared to a software equivalent implementation.
We consider the problem of creating Digital Twins (DTs) of glucose metabolism in people with Type 1 Diabetes (T1D), coupling a large-scale mean population metabolic model of glucose dynamics developed from first princ...
详细信息
We consider the problem of creating Digital Twins (DTs) of glucose metabolism in people with Type 1 Diabetes (T1D), coupling a large-scale mean population metabolic model of glucose dynamics developed from first principles, with deeplearning (DL) architectures, to map actual patient data to their virtual counterparts. For the neural network component of our proposed strategy, two models were investigated: a Long Short-Term Memory (LSTM) network and a Generative Adversarial Network (GAN). Our best model outperformed significantly the mean population model with respect to evaluation metrics (LSTM vs. metabolic simulator), expressed as median (interquartile range): MAE 35.0 (28.8, 43.8) vs. 79.7 (62.4, 115.5) [mg/dL], RMSE 44.8 (37.2, 56.2) vs. 94.9 (76.3, 128.0) [mg/dL], CNRMSE 0.85 (0.83, 0.88) vs. 0.54 (0.48, 0.72), epsilon(FIT) 0.25 (0.13, 0.40) vs. 0.00 (0.00, 0.00). We showed that the proposed physiologyinformed deeplearning approach successfully mapped real patient data to virtual subjects, with the potential to enable in-silico testing of novel therapeutic strategies on a virtual population. Copyright (C) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://***/licenses/by-nc-nd/4.0)
Numerous investigations have delved into image segmentation methodologies, encompassing techniques such as thresholding, convolution, deep neural networks, and most recently, vision transformers. However, many of thes...
详细信息
ISBN:
(纸本)9798350354119
Numerous investigations have delved into image segmentation methodologies, encompassing techniques such as thresholding, convolution, deep neural networks, and most recently, vision transformers. However, many of these studies overlook the imperative of optimizing image segmentation for various processing cores or specialized hardware like ASICs or FPGAs. This paper presents our implementation of both a neural network model and a vision transformer model based on UNET (UNETR) for image segmentation tasks, utilizing the Lapa dataset. We meticulously compare their training and prediction accuracies. Additionally, we introduce mixed-precision architectures for both models, aiming to enhance real-timeimage segmentation performance on both CPU and GPU platforms. Furthermore, we propose a novel mixed-precision FPGA architecture developed in Vivado software. This architecture is specifically tailored to optimize inference and streaming capabilities on FPGA devices. The hallmark contribution of our research lies in the development of an FPGA accelerator designed for the UNETR model. Our findings reveal substantial enhancements in the training speed of our model on GPU platforms with the application of mixed precision, as well as notable improvements in latency during real-time inference on FPGA. Notably, we achieved an F1-score and training accuracy of approximately 90% over 60 training epochs.
暂无评论