The received signal strength (RSS) fingerprint-based technique is extensively utilized for indoor localization, as it does not require time synchronization. However, conventional RSS fingerprint localization schemes r...
详细信息
This study investigates the three-dimensional dynamics of droplet splitting in bifurcation microchannels using a two-phase flow simulation with the Level Set method implemented in COMSOL Multiphysics, capturing the be...
详细信息
Agriculture plays a vital part in the economy of some countries, providing benefits such as food, national income, and employment. This results in substantial growth in agricultural produce, financial losses, and a de...
详细信息
ISBN:
(数字)9798331529246
ISBN:
(纸本)9798331529253
Agriculture plays a vital part in the economy of some countries, providing benefits such as food, national income, and employment. This results in substantial growth in agricultural produce, financial losses, and a decline in quality and quantity. Farmers have traditionally detected diseases through eye tests, but this method has drawbacks. However, previous research suggests that binary classification needs improvement because some types are more challenging to determine. To address these issues, various pepper diseases are classified based on DL techniques to address these issues. First, image pre-processing uses a Median Filter (MF) to reduce noise and increase brightness. Furthermore, the Active Contour Segmentation (ACS) method detects affected and unaffected areas in the image. Moreover, the Texture Color Intensity Feature Selection (TCIFS) method is employed to select the optimal features of pepper disease. Finally, the proposed Capsule Network with Convolutional Neural Network (CapsNet-CNN) technique detects pepper leaf disease. The proposed method attains an accuracy of 98.32%, recall of 97.43%, and precision of 98.04% compared to other methods.
Graph-based Sequential Recommender systems (GSRs) have gained significant research attention due to their ability to simultaneously handle user-item interactions and sequential relationships between items. Current GSR...
详细信息
In this work, we formulate an accelerated image fusion algorithm for Unmanned Aerial Vehicle (UAV) application which is based on image stitching using invariant features. By utilizing parallel computing techniques for...
详细信息
Current violence detection systems often face challenges such as high computational costs, real-time processing difficulties, and limited precision in detecting violent activities in dynamic environments. This researc...
详细信息
ISBN:
(数字)9798331527549
ISBN:
(纸本)9798331527556
Current violence detection systems often face challenges such as high computational costs, real-time processing difficulties, and limited precision in detecting violent activities in dynamic environments. This research addresses these issues by proposing a hybrid deep learning model that integrates MobileNetV2 for efficient spatial feature extraction and ConvLSTM (Convolutional Long Short-Term Memory) for capturing temporal patterns. The model effectively discriminates between aggressive and non-aggressive behaviours in real-time video streams. Using the Kaggle Violence Dataset, which comprises 1,000 video clips of violent and non-violent activities, the system achieved an accuracy of 96%, showcasing its high performance. Additionally, an integrated alert system enhances practical application by sending real-time notifications via a Telegram bot, including details such as location, timestamp, and camera ID. This solution ensures timely responses to violent incidents, addressing the core issues of computational efficiency and real-time precision, while offering scalability and applicability in diverse scenarios.
Quantized Spiking Neural Networks (QSNNs) offer superior energy efficiency and are well-suited for deployment on resource-limited edge devices. However, limited bit-width weight and membrane potential result in a nota...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Quantized Spiking Neural Networks (QSNNs) offer superior energy efficiency and are well-suited for deployment on resource-limited edge devices. However, limited bit-width weight and membrane potential result in a notable performance decline. In this study, we first identify a new underlying cause for this decline: the loss of historical information due to the quantized membrane potential. To tackle this issue, we introduce a memory-free quantization method that captures all historical information without directly storing membrane potentials, resulting in better performance with less memory requirements. To further improve the computational efficiency, we propose a parallel training and asynchronous inference framework that greatly increases training speed and energy efficiency. We combine the proposed memory-free quantization and parallel computation methods to develop a high-performance and efficient QSNN, named MFP-QSNN. Extensive experiments show that our MFP-QSNN achieves state-of-the-art performance on various static and neuromorphic image datasets, requiring less memory and faster training speeds. The efficiency and efficacy of the MFP-QSNN highlight its potential for energy-efficient neuromorphic computing.
Semi-supervised learning (SSL) addresses the scarcity of annotated data in medical image segmentation by leveraging unlabeled samples to enhance model training. Currently, some methods employ a co-training strategy to...
详细信息
ISBN:
(数字)9798331520526
ISBN:
(纸本)9798331520533
Semi-supervised learning (SSL) addresses the scarcity of annotated data in medical image segmentation by leveraging unlabeled samples to enhance model training. Currently, some methods employ a co-training strategy to achieve semi-supervised segmentation. However, many studies overlook the feature interaction between the two subnets. Moreover, effectively reducing the effects of inconsistent predictions from the two subnets remains challenging. In this paper, we propose a novel Discrepancy-induced Cross-subnet Interaction network for semi-supervised medical image segmentation. Specifically, we present a Cross-subnet Interaction Module (CIM) to strengthen the interaction between the two feature maps, which helps reduce model cognitive bias. Additionally, we propose a Discrepancy-induced Weight Loss (DWL) mechanism to focus the model on regions with inconsistent predictions from the two subnets, thereby mitigating the effects of these inconsistencies. Experimental results on three segmentation tasks demonstrate that the proposed model achieves superior performance compared to existing semi-supervised segmentation methods.
Post-Training Quantization (PTQ) is pivotal for deploying large language models (LLMs) within resource-limited settings by significantly reducing resource demands. However, existing PTQ strategies underperform at low ...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Post-Training Quantization (PTQ) is pivotal for deploying large language models (LLMs) within resource-limited settings by significantly reducing resource demands. However, existing PTQ strategies underperform at low bit levels (< 3 bits) due to the significant difference between the quantized and original weights. To enhance the quantization performance at low bit widths, we introduce a Mixed-precision Graph Neural PTQ (MG-PTQ) approach, employing a graph neural network (GNN) module to capture dependencies among weights and adaptively assign quantization bit-widths. Through the information propagation of the GNN module, our method more effectively captures dependencies among target weights, leading to a more accurate assessment of weight importance and optimized allocation of quantization strategies. Extensive experiments on the WikiText2 and C4 datasets demonstrate that our MG-PTQ method outperforms previous state-of-the-art PTQ method GPTQ, setting new benchmarks for quantization performance under low-bit (< 3 bits) conditions.
Process-based learning is crucial for the transmission of intangible cultural heritage, especially in complex arts like Chinese calligraphy, where mastering techniques cannot be achieved by merely observing the final ...
详细信息
暂无评论