Search

Liposome-based anticancer agents take advantage of the increased vascular permeability and transvascular pressure gradients for selective accumulation in tumors, a phenomenon known as the enhanced permeability and retention(EPR) effect. The EPR effect has motivated the clinical use of nano-therapeutics, with mixed results on treatment outcome. High interstitial fluid pressure (IFP) has been shown to limit liposome drug delivery to central tumour regions. Furthermore, high IFP is an independent prognostic biomarker for treatment efficacy in radiation therapy and chemotherapy for some solid cancers. Therefore, accurately measuring spatial liposome accumulation and IFP distribution within a solid tumour is crucial for optimal treatment planning. In this paper, we develop a model capable of predicting voxel-by-voxel intratumoral liposome accumulation and IFP using pre and post administration imaging. Our approach is based on physics informed machine learning, a novel technique combining machine learning and partial differential equations. through application to a set of mouse data and a set of synthetically-generated tumours, we show that our approach accurately predicts the spatial liposome accumulation and IFP for an individual tumour while relying on minimal information. This is an important result with applications for forecasting tumour progression and designing treatment. [ABSTRACT FROM AUTHOR]

Pancreatic fine-needle aspirations are the gold-standard diagnostic procedure for the evaluation of pancreatic ductal adenocarcinoma. A suspicion for malignancy can escalate towards chemotherapy followed by a major surgery and therefore is a high-stakes task for the pathologist. In this paper, we propose a deep learning framework, MIPCL, that can serve as a helpful screening tool, predicting the presence or absence of cancer. We also reproduce two deep learning models that have found success in surgical pathology for our cytopathology study. Our MIPCL significantly improves over both models across all evaluated metrics (F1-Score: 87.97% vs 88.70% vs 91.07%; AUROC: 0.9159 vs. 0.9051 vs 0.9435). Additionally, our model is able to recover the most contributing regions on the slide for the final prediction. We also present a dataset curation strategy that increases the number of training examples from an existing dataset, thereby reducing the resource burden tied to collecting and scanning additional cases. [ABSTRACT FROM AUTHOR]

Bats are a crucial component within ecosystems, providing valuable ecosystem services such as pollination and pest control. In practical conservation efforts, the classification and identification of bats are essential in order to develop effective conservation management programs for bats and their habitats. Traditionally, the identification of bats has been a manual and time-consuming process. With the development of artificial intelligence technology, the accuracy and speed of identification work of such fine-grained images as bats identification can be greatly improved. Bats identification relies on the fine features of their beaks and faces, so mining the fine-grained information in images is crucial to improve the accuracy of bats identification. This paper presents a deep learning-based model designed for the rapid and precise identification of common horseshoe bats (Chiroptera: Rhinolophidae: Rhinolophus) from Southern China. The model was developed by utilizing a comprehensive dataset of 883 high-resolution images of seven distinct Rhinolophus species which were collected during surveys conducted between 2010 and 2022. An improved EfficientNet model with an attention mechanism module is architected to mine the fine-grained appearance of these Rhinolophus. The performance of the model beat other classical models, including SqueezeNet, AlexNet, VGG16_BN, ShuffleNetV2, GoogleNet, ResNet50 and EfficientNet_B0, according to the predicting precision, recall, accuracy, F1-score. Our model achieved the highest identification accuracy of 94.22% and an F1-score of 0.948 with low computational complexity. Heat maps obtained with Grad-CAM show that our model meets the identification criteria of the morphology of Rhinolophus. Our study highlights the potential of artificial intelligence technology for the identification of small mammals, and facilitating fast species identification in the future. [ABSTRACT FROM AUTHOR]

This work proposes a new machine learning (ML)-based paradigm aiming to enhance the computational efficiency of non-equilibrium reacting flow simulations while ensuring compliance with the underlying physics. The framework combines dimensionality reduction and neural operators through a hierarchical and adaptive deep learning strategy to learn the solution of multi-scale coarse-grained governing equations for chemical kinetics. The proposed surrogate's architecture is structured as a tree, with leaf nodes representing separate neural operator blocks where physics is embedded in the form of multiple soft and hard constraints. The hierarchical attribute has two advantages: (i) It allows the simplification of the training phase via transfer learning, starting from the slowest temporal scales; (ii) It accelerates the prediction step by enabling adaptivity as the surrogate's evaluation is limited to the necessary leaf nodes based on the local degree of non-equilibrium of the gas. The model is applied to the study of chemical kinetics relevant for application to hypersonic flight, and it is tested here on pure oxygen gas mixtures. In 0- D scenarios, the proposed ML framework can adaptively predict the dynamics of almost thirty species with a maximum relative error of 4.5% for a wide range of initial conditions. Furthermore, when employed in 1- D shock simulations, the approach shows accuracy ranging from 1% to 4.5% and a speedup of one order of magnitude compared to conventional implicit schemes employed in an operator-splitting integration framework. Given the results presented in the paper, this work lays the foundation for constructing an efficient ML-based surrogate coupled with reactive Navier-Stokes solvers for accurately characterizing non-equilibrium phenomena in multi-dimensional computational fluid dynamics simulations. [ABSTRACT FROM AUTHOR]

Distributed learning, as the most popular solution for training large-scale data for deep learning, consists of multiple participants collaborating on data training tasks. However, the malicious behavior of some during the training process, like Byzantine participants who would interrupt or control the learning process, will trigger the crisis of data security. Although recent existing defense mechanisms use the variability of Byzantine node gradients to clear Byzantine values, it is still unable to identify and then clear the delicate disturbance/attack. To address this critical issue, we propose an algorithm named consensus aggregation in this paper. This algorithm allows computational nodes to use the information of verification nodes to verify the effectiveness of the gradient in the perturbation attack, reaching a consensus based on the effective verification of the gradient. Then the server node uses the gradient as the valid gradient for gradient aggregation calculation through the consensus reached by other computing nodes. On the MNIST and CIFAR10 datasets, when faced with Drift attacks, the proposed algorithm outperforms common existing aggregation algorithms (Krum, Trimmed Mean, Bulyan), with accuracies of 93.3%, 94.06% (MNIST dataset), and 48.66%, 51.55% (CIFAR10 dataset), respectively. This is an improvement of 3.0%, 3.8% (MNIST dataset), and 19.0%, 26.1% (CIFAR10 dataset) over the current state-of-the-art methods, and successfully defended against other attack methods. [ABSTRACT FROM AUTHOR]

Automated multi-organ segmentation plays an essential part in the computer-aided diagnostic (CAD) of chest X-ray fluoroscopy. However, developing a CAD system for the anatomical structure segmentation remains challenging due to several indistinct structures, variations in the anatomical structure shape among different individuals, the presence of medical tools, such as pacemakers and catheters, and various artifacts in the chest radiographic images. In this paper, we propose a robust deep learning segmentation framework for the anatomical structure in chest radiographs that utilizes a dual encoder–decoder convolutional neural network (CNN). The first network in the dual encoder–decoder structure effectively utilizes a pre-trained VGG19 as an encoder for the segmentation task. The pre-trained encoder output is fed into the squeeze-and-excitation (SE) to boost the network's representation power, which enables it to perform dynamic channel-wise feature calibrations. The calibrated features are efficiently passed into the first decoder to generate the mask. We integrated the generated mask with the input image and passed it through a second encoder–decoder network with the recurrent residual blocks and an attention the gate module to capture the additional contextual features and improve the segmentation of the smaller regions. Three public chest X-ray datasets are used to evaluate the proposed method for multi-organs segmentation, such as the heart, lungs, and clavicles, and single-organ segmentation, which include only lungs. The results from the experiment show that our proposed technique outperformed the existing multi-class and single-class segmentation methods. [ABSTRACT FROM AUTHOR]

Road manhole covers are crucial components of urban infrastructure; however, inadequate maintenance or poor marking can pose safety risks to vehicular traffic. This paper presents a method for detecting road manhole covers using a stereo depth camera and the MGB-YOLO model. We curated a robust image dataset and performed image enhancement and annotation. The MGB-YOLO model was developed by optimizing the YOLOv5s network with MobileNet-V3, Global Attention Mechanism (GAM), and BottleneckCSP, striking a balance between detection accuracy and model efficiency. Our method achieved an impressive accuracy of 96.6%, surpassing the performance of Faster RCNN, SSD, YOLOv5s, YOLOv7 and YOLOv8s models with an increased mean average precision (mAP) of 15.6%, 6.9%, 0.7%, 0.5% and 0.5%, respectively. Additionally, we have reduced the model's size and the number of parameters, making it highly suitable for deployment on in-vehicle embedded devices. These results underscore the effectiveness of our approach in detecting road manhole covers, offering valuable insights for vehicle-based manhole cover detection and contributing to the reduction of accidents and enhanced driving comfort. [ABSTRACT FROM AUTHOR]

Surface-enhanced Raman spectroscopy (SERS), as a rapid, non-invasive and reliable spectroscopic detection technique, has promising applications in disease screening and diagnosis. In this paper, an annealed silver nanoparticles/porous silicon Bragg reflector (AgNPs/PSB) composite SERS substrate with high sensitivity and strong stability was prepared by immersion plating and heat treatment using porous silicon Bragg reflector (PSB) as the substrate. The substrate combines the five deep learning algorithms of the improved AlexNet, ResNet, SqueezeNet, temporal convolutional network (TCN) and multiscale fusion convolutional neural network (MCNN). We constructed rapid screening models for patients with primary Sjögren's syndrome (pSS) and healthy controls (HC), diabetic nephropathy patients (DN) and healthy controls (HC), respectively. The results showed that the annealed AgNPs/PSB composite SERS substrates performed well in diagnosing. Among them, the MCNN model had the best classification effect in the two groups of experiments, with an accuracy rate of 94.7% and 92.0%, respectively. Previous studies have indicated that the AgNPs/PSB composite SERS substrate, combined with machine learning algorithms, has achieved promising classification results in disease diagnosis. This study shows that SERS technology based on annealed AgNPs/PSB composite substrate combined with deep learning algorithm has a greater developmental prospect and research value in the early identification and screening of immune diseases and chronic kidney disease, providing reference ideas for non-invasive and rapid clinical medical diagnosis of patients. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) is one of the main causes of blindness in people around the world. Early diagnosis and treatment of DR can be accomplished by organizing large regular screening programs. Still, it is difficult to spot diabetic retinopathy timely because the situation might not indicate signs in the primary stages of the disease. Due to a drastic increase in diabetic patients, there is an urgent need for efficient diabetic retinopathy detecting systems. Auto-encoders, sparse coding, and limited Boltzmann machines were used as a few past deep learning (DL) techniques and features for the classification of DR. Convolutional Neural Networks (CNN) have been identified as a promising solution for detecting and classifying DR. We employ the deep learning capabilities of efficient net batch normalization (BNs) pre-trained models to automatically acquire discriminative features from fundus images. However, we successfully achieved F1 scores above 80% on all efficient net BNs in the EYE-PACS dataset (calculated F1 score for DeepDRiD another dataset) and the results are better than previous studies. In this paper, we improved the accuracy and F1 score of the efficient net BNs pre-trained models on the EYE-PACS dataset by applying a Gaussian Smooth filter and data augmentation transforms. Using our proposed technique, we have achieved F1 scores of 84% and 87% for EYE-PACS and DeepDRiD. [ABSTRACT FROM AUTHOR]

Sentence-level sentiment analysis (SLSA) aims to identify the overall sentiment polarity conveyed in a given sentence. The state-of-the-art performance of SLSA has been achieved by deep learning models. However, depending on the i.i.d (independent and identically distributed) assumption, the performance of these deep learning models may fall short in real scenarios, where the distributions of training and target data are almost certainly different to some extent. In this paper, we propose a supervised solution based on the non-i.i.d paradigm of gradual machine learning (GML) for SLSA. It begins with some labeled observations, and gradually labels target instances in the order of increasing hardness by iterative knowledge conveyance. It leverages labeled samples for supervised deep feature extraction, and constructs a factor graph based on the extracted features to enable gradual knowledge conveyance. Specifically, it employs a polarity classifier to detect polarity similarity between close neighbors in an embedding space, and a separate binary semantic network to extract implicit polarity relations between arbitrary instances. Our extensive experiments on benchmark datasets show that the proposed approach achieves the state-of-the-art performance on all benchmark datasets. Our work clearly demonstrates that by leveraging DNN for feature extraction, GML can easily outperform the pure DNN solutions. [ABSTRACT FROM AUTHOR]

Air pollution is a leading cause of human diseases. Accurate air quality predictions are critical to human health. However, it is difficult to extract spatiotemporal features among complex spatiotemporal dependencies effectively. Most existing methods focus on constructing multiple spatial dependencies and ignore the systematic analysis of spatial dependencies. We found that besides spatial proximity stations, functional similarity stations, and temporal pattern similarity stations, the shared spatial dependencies also exist in the complete spatial dependencies. In this paper, we propose a novel deep learning model, the spatiotemporal adaptive attention graph convolution model, for city-level air quality prediction, in which the prediction of future short-term series of PM2.5 readings is preferred. Specifically, we encode multiple spatiotemporal dependencies and construct complete spatiotemporal interactions between stations using station-level attention. Among them, we design a Bi-level sharing strategy to extract shared inter-station relationship features between certain stations efficiently. Then we extract multiple spatiotemporal features with multiple decoders, which it is extracted from the complete spatial dependencies between stations. Finally, we fuse multiple spatiotemporal features with a gating mechanism for multi-step predictions. Our model achieves state-of-the-art experimental results in several real-world datasets. [ABSTRACT FROM AUTHOR]

Aphid infestation poses a significant threat to crop production, rural communities, and global food security. While chemical pest control is crucial for maximizing yields, applying chemicals across entire fields is both environmentally unsustainable and costly. Hence, precise localization and management of aphids are essential for targeted pesticide application. The paper primarily focuses on using deep learning models for detecting aphid clusters. We propose a novel approach for estimating infection levels by detecting aphid clusters. To facilitate this research, we have captured a large-scale dataset from sorghum fields, manually selected 5447 images containing aphids, and annotated each individual aphid cluster within these images. To facilitate the use of machine learning models, we further process the images by cropping them into patches, resulting in a labeled dataset comprising 151,380 image patches. Then, we implemented and compared the performance of four state-of-the-art object detection models (VFNet, GFLV2, PAA, and ATSS) on the aphid dataset. Extensive experimental results show that all models yield stable similar performance in terms of average precision and recall. We then propose to merge close neighboring clusters and remove tiny clusters caused by cropping, and the performance is further boosted by around 17%. The study demonstrates the feasibility of automatically detecting and managing insects using machine learning models. The labeled dataset will be made openly available to the research community. [ABSTRACT FROM AUTHOR]

Based on the deep learning method, a network model that can quickly and accurately identify the species of Fritillaria cirrhosa species was constructed. The learning method based on deep residual convolutional neural network was used to input the unprocessed original image directly as input, and the features of the image were extracted through convolution and pooling operations. On this basis, the ResNet34 model was improved, and the additional fully connected layer was added in front of the Softmax classifier to improve the learning ability of the network model. Total of 3915 images of three kinds of Fritillaria cirrhosa were used as data sources for the experiments, among which 160 images of each type were randomly selected to form the validation set. The final training set recognition accuracy rate was 95.8%, the validation set accuracy rate reached 92.3%, and the test set accuracy rate was 88.7%. The image recognition method of Fritillaria cirrhosa based on deep learning proposed in this paper is effective and feasible, which can quickly and accurately identify the species of Fritillaria cirrhosa species, and provides a new idea for the intelligent recognition of Chinese medicinal materials. [ABSTRACT FROM AUTHOR]

The world's population is expected to exceed 9 billion people by 2050, necessitating a 70% increase in agricultural output and food production to meet the demand. Due to resource shortages, climate change, the COVID-19 pandemic, and highly harsh socioeconomic predictions, such a demand is challenging to complete without using computation and forecasting methods. Machine learning has grown with big data and high-performance computers technologies to open up new data-intensive scientific opportunities in the multidisciplinary agri-technology area. Throughout the plant's developmental period, diseases and pests are natural disasters, from seed production to seedling growth. This paper introduces an early diagnosis framework for plant diseases based on fog computing and edge environment by IoT sensors measurements and communication technologies. The effectiveness of employing pre-trained CNN architectures as feature extractors in identifying plant illnesses has been studied. As feature extractors, standard pre-trained CNN models, AlexNet are employed. The obtained in-depth features are eliminated by proposing a revised version of the grey wolf optimization (GWO) algorithm that approved its efficiency through experiments. The features subset selected were used to train the SVM classifier. Ten datasets for different plants are utilized to assess the proposed model. According to the findings, the proposed model achieved better outcomes for all used datasets. As an average for all datasets, the accuracy of the proposed model is 93.84 compared to 85.49, 87.89, 87.04 for AlexNet, GoogleNet, and the SVM, respectively. [ABSTRACT FROM AUTHOR]

To assist ophthalmologists in diagnosing retinal abnormalities, Computer Aided Diagnosis has played a significant role. In this paper, a particular Convolutional Neural Network based on Wavelet Scattering Transform (WST) is used to detect one to four retinal abnormalities from Optical Coherence Tomography (OCT) images. Predefined wavelet filters in this network decrease the computation complexity and processing time compared to deep learning methods. We use two layers of the WST network to obtain a direct and efficient model. WST generates a sparse representation of the images which is translation-invariant and stable concerning local deformations. Next, a Principal Component Analysis classifies the extracted features. We evaluate the model using four publicly available datasets to have a comprehensive comparison with the literature. The accuracies of classifying the OCT images of the OCTID dataset into two and five classes were 100 % and 82.5 % , respectively. We achieved an accuracy of 96.6 % in detecting Diabetic Macular Edema from Normal ones using the TOPCON device-based dataset. Heidelberg and Duke datasets contain DME, Age-related Macular Degeneration, and Normal classes, in which we achieved accuracy of 97.1 % and 94.4 % , respectively. A comparison of our results with the state-of-the-art models shows that our model outperforms these models for some assessments or achieves nearly the best results reported so far while having a much smaller computational complexity. [ABSTRACT FROM AUTHOR]

Pancreatic fine-needle aspirations are the gold-standard diagnostic procedure for the evaluation of pancreatic ductal adenocarcinoma. A suspicion for malignancy can escalate towards chemotherapy followed by a major surgery and therefore is a high-stakes task for the pathologist. In this paper, we propose a deep learning framework, MIPCL, that can serve as a helpful screening tool, predicting the presence or absence of cancer. We also reproduce two deep learning models that have found success in surgical pathology for our cytopathology study. Our MIPCL significantly improves over both models across all evaluated metrics (F1-Score: 87.97% vs 88.70% vs 91.07%; AUROC: 0.9159 vs. 0.9051 vs 0.9435). Additionally, our model is able to recover the most contributing regions on the slide for the final prediction. We also present a dataset curation strategy that increases the number of training examples from an existing dataset, thereby reducing the resource burden tied to collecting and scanning additional cases. [ABSTRACT FROM AUTHOR]

Human activity recognition (HAR) is one of the key applications of health monitoring that requires continuous use of wearable devices to track daily activities. The most efficient supervised machine learning (ML)-based approaches for predicting human activity are based on a continuous stream of sensor data. Sensor data analysis for human activity recognition using conventional algorithms and deep learning (DL) models shows promising results, but evaluating their ambiguity in decision-making is still challenging. In order to solve these issues, the paper proposes a novel Wasserstein gradient flow legonet WGF-LN-based human activity recognition system. At first, the input data is pre-processed. From the pre-processed data, the features are extracted using Haar Wavelet mother- Symlet wavelet coefficient scattering feature extraction (HS-WSFE). After that, the interest features are selected from the extracted features using (Binomial Distribution integrated-Golden Eagle Optimization) BD-GEO. The important features are then post-processed using the scatter plot matrix method. Obtained post-processing features are finally given into the WGF-LN for classifying human activities. From these experiments, the results can be obtained and showed the efficacy of the proposed model. [ABSTRACT FROM AUTHOR]

In this paper, we present a channel estimation approach based on deep learning to solve the problem that the orthogonal frequency division multiplexing (OFDM) system channel estimation algorithm cannot accurately obtain the channel state information in the complex environment of the mine, resulting in system performance degradation. First, LS channel estimation matrix is considered as a low-resolution image and the actual channel state information is considered as a high-resolution image. Then the optimization of the LS channel estimation matrix is achieved by the FSRCNN image super-resolution algorithm. We validate the effectiveness of the proposed algorithm by conducting experiments in different channel environments, different number of pilots, and mismatched signal-to-noise ratio scenarios. The simulation results show that the proposed scheme is much better than the traditional LS channel estimation method and the DFT-LS channel estimation method, and the accuracy of the proposed scheme approaches that of the MMSE channel estimation method when the number of pilots is low. [ABSTRACT FROM AUTHOR]

Pancreatic fine-needle aspirations are the gold-standard diagnostic procedure for the evaluation of pancreatic ductal adenocarcinoma. A suspicion for malignancy can escalate towards chemotherapy followed by a major surgery and therefore is a high-stakes task for the pathologist. In this paper, we propose a deep learning framework, MIPCL, that can serve as a helpful screening tool, predicting the presence or absence of cancer. We also reproduce two deep learning models that have found success in surgical pathology for our cytopathology study. Our MIPCL significantly improves over both models across all evaluated metrics (F1-Score: 87.97% vs 88.70% vs 91.07%; AUROC: 0.9159 vs. 0.9051 vs 0.9435). Additionally, our model is able to recover the most contributing regions on the slide for the final prediction. We also present a dataset curation strategy that increases the number of training examples from an existing dataset, thereby reducing the resource burden tied to collecting and scanning additional cases. [ABSTRACT FROM AUTHOR]

Adding hooked industrial steel fibers (ISF) to concrete boosts its tensile and flexural strength. However, the understanding of ISF's influence on the compressive strength (CS) behavior of concrete is still questioned by the scientific society. The presented paper aims to use machine learning (ML) and deep learning (DL) algorithms to predict the CS of steel fiber reinforced concrete (SFRC) incorporating hooked ISF based on the data collected from the open literature. Accordingly, 176 sets of data are collected from different journals and conference papers. Based upon the initial sensitivity analysis, the most influential parameters like water-to-cement (W/C) ratio and content of fine aggregates (FA) tend to decrease the CS of SFRC. Meanwhile, the CS of SFRC could be enhanced by increasing the amount of superplasticizer (SP), fly ash, and cement (C). The least contributing factors include the maximum size of aggregates (Dmax) and the length-to-diameter ratio of hooked ISFs (L/DISF). Several statistical parameters are also used as metrics to evaluate the performance of implemented models, such as coefficient of determination (R2), mean absolute error (MAE), and mean of squared error (MSE). Among different ML algorithms, convolutional neural network (CNN) with R2 = 0.928, RMSE = 5.043, and MAE = 3.833 shows higher accuracy. On the other hand, K-nearest neighbor (KNN) algorithm with R2 = 0.881, RMSE = 6.477, and MAE = 4.648 results in the weakest performance. [ABSTRACT FROM AUTHOR]

The tubule index is a vital prognostic measure in breast cancer tumor grading and is visually evaluated by pathologists. In this paper, a computer-aided patch-based deep learning tubule segmentation framework, named Tubule-U-Net, is developed and proposed to segment tubules in Whole Slide Images (WSI) of breast cancer. Moreover, this paper presents a new tubule segmentation dataset consisting of 30820 polygonal annotated tubules in 8225 patches. The Tubule-U-Net framework first uses a patch enhancement technique such as reflection or mirror padding and then employs an asymmetric encoder-decoder semantic segmentation model. The encoder is developed in the model by various deep learning architectures such as EfficientNetB3, ResNet34, and DenseNet161, whereas the decoder is similar to U-Net. Thus, three different models are obtained, which are EfficientNetB3-U-Net, ResNet34-U-Net, and DenseNet161-U-Net. The proposed framework with three different models, U-Net, U-Net++, and Trans-U-Net segmentation methods are trained on the created dataset and tested on five different WSIs. The experimental results demonstrate that the proposed framework with the EfficientNetB3 model trained on patches obtained using the reflection padding and tested on patches with overlapping provides the best segmentation results on the test data and achieves 95.33%, 93.74%, and 90.02%, dice, recall, and specificity scores, respectively. [ABSTRACT FROM AUTHOR]

Historical documents contain essential information about the past, including places, people, or events. Many of these valuable cultural artifacts cannot be further examined due to aging or external influences, as they are too fragile to be opened or turned over, so their rich contents remain hidden. Terahertz (THz) imaging is a nondestructive 3D imaging technique that can be used to reveal the hidden contents without damaging the documents. As noise or imaging artifacts are predominantly present in reconstructed images processed by standard THz reconstruction algorithms, this work intends to improve THz image quality with deep learning. To overcome the data scarcity problem in training a supervised deep learning model, an unsupervised deep learning network (CycleGAN) is first applied to generate paired noisy THz images from clean images (clean images are generated by a handwriting generator). With such synthetic noisy-to-clean paired images, a supervised deep learning model using Pix2pixGAN is trained, which is effective to enhance real noisy THz images. After Pix2pixGAN denoising, 99% characters written on one-side of the Xuan paper can be clearly recognized, while 61% characters written on one-side of the standard paper are sufficiently recognized. The average perceptual indices of Pix2pixGAN processed images are 16.83, which is very close to the average perceptual index 16.19 of clean handwriting images. Our work has important value for THz-imaging-based nondestructive historical document analysis. [ABSTRACT FROM AUTHOR]

Natural hazard named entity recognition is a technique used to recognize natural hazard entities from a large number of texts. The method of natural hazard named entity recognition can facilitate acquisition of natural hazards information and provide reference for natural hazard mitigation. The method of named entity recognition has many challenges, such as fast change, multiple types and various forms of named entities. This can introduce difficulties in research of natural hazard named entity recognition. To address the above problem, this paper constructed a natural disaster annotated corpus for training and evaluation model, and selected and compared several deep learning methods based on word vector features. A deep learning method for natural hazard named entity recognition can automatically mine text features and reduce the dependence on manual rules. This paper compares and analyzes the deep learning models from three aspects: pretraining, feature extraction and decoding. A natural hazard named entity recognition method based on deep learning is proposed, namely XLNet-BiLSTM-CRF model. Finally, the research hotspots of natural hazards papers in the past 10 years were obtained through this model. After training, the precision of the XLNet-BilSTM-CRF model is 92.80%, the recall rate is 91.74%, and the F1-score is 92.27%. The results show that this method, which is superior to other methods, can effectively recognize natural hazard named entities. [ABSTRACT FROM AUTHOR]

Road manhole covers are crucial components of urban infrastructure; however, inadequate maintenance or poor marking can pose safety risks to vehicular traffic. This paper presents a method for detecting road manhole covers using a stereo depth camera and the MGB-YOLO model. We curated a robust image dataset and performed image enhancement and annotation. The MGB-YOLO model was developed by optimizing the YOLOv5s network with MobileNet-V3, Global Attention Mechanism (GAM), and BottleneckCSP, striking a balance between detection accuracy and model efficiency. Our method achieved an impressive accuracy of 96.6%, surpassing the performance of Faster RCNN, SSD, YOLOv5s, YOLOv7 and YOLOv8s models with an increased mean average precision (mAP) of 15.6%, 6.9%, 0.7%, 0.5% and 0.5%, respectively. Additionally, we have reduced the model's size and the number of parameters, making it highly suitable for deployment on in-vehicle embedded devices. These results underscore the effectiveness of our approach in detecting road manhole covers, offering valuable insights for vehicle-based manhole cover detection and contributing to the reduction of accidents and enhanced driving comfort. [ABSTRACT FROM AUTHOR]

Bats are a crucial component within ecosystems, providing valuable ecosystem services such as pollination and pest control. In practical conservation efforts, the classification and identification of bats are essential in order to develop effective conservation management programs for bats and their habitats. Traditionally, the identification of bats has been a manual and time-consuming process. With the development of artificial intelligence technology, the accuracy and speed of identification work of such fine-grained images as bats identification can be greatly improved. Bats identification relies on the fine features of their beaks and faces, so mining the fine-grained information in images is crucial to improve the accuracy of bats identification. This paper presents a deep learning-based model designed for the rapid and precise identification of common horseshoe bats (Chiroptera: Rhinolophidae: Rhinolophus) from Southern China. The model was developed by utilizing a comprehensive dataset of 883 high-resolution images of seven distinct Rhinolophus species which were collected during surveys conducted between 2010 and 2022. An improved EfficientNet model with an attention mechanism module is architected to mine the fine-grained appearance of these Rhinolophus. The performance of the model beat other classical models, including SqueezeNet, AlexNet, VGG16_BN, ShuffleNetV2, GoogleNet, ResNet50 and EfficientNet_B0, according to the predicting precision, recall, accuracy, F1-score. Our model achieved the highest identification accuracy of 94.22% and an F1-score of 0.948 with low computational complexity. Heat maps obtained with Grad-CAM show that our model meets the identification criteria of the morphology of Rhinolophus. Our study highlights the potential of artificial intelligence technology for the identification of small mammals, and facilitating fast species identification in the future. [ABSTRACT FROM AUTHOR]

Human monkeypox is a very unusual virus that can devastate society. Early identification and diagnosis are essential to treat and manage an illness effectively. Human monkeypox disease detection using deep learning models has attracted increasing attention recently. The virus that causes monkeypox may be passed to people, making it a zoonotic illness. The latest monkeypox epidemic has hit more than 40 nations. Computer-assisted approaches using Deep Learning techniques for automatically identifying skin lesions have shown to be a viable alternative in light of the fast proliferation and ever-growing problems of supplying PCR (Polymerase Chain Reaction) Testing in places with limited availability. In this research, we introduce a deep learning model for detecting human monkeypoxes that is accurate and resilient by tuning its hyper-parameters. We employed a mixture of convolutional neural networks and transfer learning strategies to extract characteristics from medical photos and properly identify them. We also used hyperparameter optimization strategies to fine-tune the Model and get the best possible results. This paper proposes a Yolov5 model-based method for differentiating between chickenpox and Monkeypox lesions on skin pictures. The Roboflow skin lesion picture dataset was subjected to three different hyperparameter tuning strategies: the SDG optimizer, the Bayesian optimizer, and Learning without Forgetting. The proposed Model had the highest classification accuracy (98.18%) when applied to photos of monkeypox skin lesions. Our findings show that the suggested Model surpasses the current best-in-class models and may be used in clinical settings for actual Human Monkeypox disease detection and diagnosis. [ABSTRACT FROM AUTHOR]

Surface-enhanced Raman spectroscopy (SERS), as a rapid, non-invasive and reliable spectroscopic detection technique, has promising applications in disease screening and diagnosis. In this paper, an annealed silver nanoparticles/porous silicon Bragg reflector (AgNPs/PSB) composite SERS substrate with high sensitivity and strong stability was prepared by immersion plating and heat treatment using porous silicon Bragg reflector (PSB) as the substrate. The substrate combines the five deep learning algorithms of the improved AlexNet, ResNet, SqueezeNet, temporal convolutional network (TCN) and multiscale fusion convolutional neural network (MCNN). We constructed rapid screening models for patients with primary Sjögren's syndrome (pSS) and healthy controls (HC), diabetic nephropathy patients (DN) and healthy controls (HC), respectively. The results showed that the annealed AgNPs/PSB composite SERS substrates performed well in diagnosing. Among them, the MCNN model had the best classification effect in the two groups of experiments, with an accuracy rate of 94.7% and 92.0%, respectively. Previous studies have indicated that the AgNPs/PSB composite SERS substrate, combined with machine learning algorithms, has achieved promising classification results in disease diagnosis. This study shows that SERS technology based on annealed AgNPs/PSB composite substrate combined with deep learning algorithm has a greater developmental prospect and research value in the early identification and screening of immune diseases and chronic kidney disease, providing reference ideas for non-invasive and rapid clinical medical diagnosis of patients. [ABSTRACT FROM AUTHOR]

An increasing and aging patient population poses a growing burden on healthcare professionals. Automation of medical imaging diagnostics holds promise for enhancing patient care and reducing manpower required to accommodate an increasing patient-population. Deep learning, a subset of machine learning, has the potential to facilitate automated diagnostics, but commonly requires large-scaled labeled datasets. In medical domains, data is often abundant but labeling is a laborious and costly task. Active learning provides a method to optimize the selection of unlabeled samples that are most suitable for improvement of the model and incorporate them into the model training process. This approach proves beneficial when only a small number of labeled samples are available. Various selection methods currently exist, but most of them employ fixed querying schedules. There is limited research on how the timing of a query can impact performance in relation to the number of queried samples. This paper proposes a novel approach called dynamic querying, which aims to optimize the timing of queries to enhance model development while utilizing as few labeled images as possible. The performance of the proposed model is compared to a model trained utilizing a fully-supervised training method, and its effectiveness is assessed based on dataset size requirements and loss rates. Dynamic querying demonstrates a considerably faster learning curve in relation to the number of labeled samples used, achieving an accuracy of 70% using only 24 samples, compared to 82% for a fully-supervised model trained on the complete training dataset of 1017 images. [ABSTRACT FROM AUTHOR]

This work proposes a new machine learning (ML)-based paradigm aiming to enhance the computational efficiency of non-equilibrium reacting flow simulations while ensuring compliance with the underlying physics. The framework combines dimensionality reduction and neural operators through a hierarchical and adaptive deep learning strategy to learn the solution of multi-scale coarse-grained governing equations for chemical kinetics. The proposed surrogate's architecture is structured as a tree, with leaf nodes representing separate neural operator blocks where physics is embedded in the form of multiple soft and hard constraints. The hierarchical attribute has two advantages: (i) It allows the simplification of the training phase via transfer learning, starting from the slowest temporal scales; (ii) It accelerates the prediction step by enabling adaptivity as the surrogate's evaluation is limited to the necessary leaf nodes based on the local degree of non-equilibrium of the gas. The model is applied to the study of chemical kinetics relevant for application to hypersonic flight, and it is tested here on pure oxygen gas mixtures. In 0- D scenarios, the proposed ML framework can adaptively predict the dynamics of almost thirty species with a maximum relative error of 4.5% for a wide range of initial conditions. Furthermore, when employed in 1- D shock simulations, the approach shows accuracy ranging from 1% to 4.5% and a speedup of one order of magnitude compared to conventional implicit schemes employed in an operator-splitting integration framework. Given the results presented in the paper, this work lays the foundation for constructing an efficient ML-based surrogate coupled with reactive Navier-Stokes solvers for accurately characterizing non-equilibrium phenomena in multi-dimensional computational fluid dynamics simulations. [ABSTRACT FROM AUTHOR]

Currently, a noteworthy volume of information is available and shared every day through participation and communication of individuals on social media. These enormous contents with the right exploit and research leads to valuable discoveries. In this study, a deep framework of learning accurate detection of women's depression is proposed. It is beneficially guided by social media content of individual posts and tweets and an essential support from psycho-linguistic for providing the indicator depression signs vocabulary that creates the embedding words necessary for building the applied approach. The presented model is validated using dual datasets extracted from Twitter: the first dataset is general data formed by 700 women from different countries; the second contains only 80 women from KSA. A third benchmark dataset CLPsych 2015 is used for comparative analysis purposes. The model proved its performance on the three datasets and the obtained and reported in this paper results shows its effectiveness. [ABSTRACT FROM AUTHOR]

Smart monitoring and surveillance systems have become one of the fundamental areas in the context of security applications in Smart Cities. In particular, video surveillance for Human Activity Recognition (HAR) applied to the recognition of potential offenders and to the detection and prevention of violent acts is a challenging task that is still undergoing. This paper presents a method based on deep learning for face recognition at a distance for security applications. Due to the absence of available datasets on face recognition at a distance, a methodology to generate a reliable dataset that relates the distance of the individuals from the camera, the focal length of the image sensors and the size in pixels of the target face is introduced. To generate the extended dataset, the Georgia Tech Face and Quality Dataset for Distance Faces databases were chosen. Our method is then tested and applied to a set of commercial image sensors for surveillance cameras using this dataset. The system achieves an average accuracy above 99% for several sensors and allows to calculate the maximum distance for a sensor to get the required accuracy in the recognition, which could be crucial in security applications in smart cities. [ABSTRACT FROM AUTHOR]

Sentence-level sentiment analysis (SLSA) aims to identify the overall sentiment polarity conveyed in a given sentence. The state-of-the-art performance of SLSA has been achieved by deep learning models. However, depending on the i.i.d (independent and identically distributed) assumption, the performance of these deep learning models may fall short in real scenarios, where the distributions of training and target data are almost certainly different to some extent. In this paper, we propose a supervised solution based on the non-i.i.d paradigm of gradual machine learning (GML) for SLSA. It begins with some labeled observations, and gradually labels target instances in the order of increasing hardness by iterative knowledge conveyance. It leverages labeled samples for supervised deep feature extraction, and constructs a factor graph based on the extracted features to enable gradual knowledge conveyance. Specifically, it employs a polarity classifier to detect polarity similarity between close neighbors in an embedding space, and a separate binary semantic network to extract implicit polarity relations between arbitrary instances. Our extensive experiments on benchmark datasets show that the proposed approach achieves the state-of-the-art performance on all benchmark datasets. Our work clearly demonstrates that by leveraging DNN for feature extraction, GML can easily outperform the pure DNN solutions. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) is one of the main causes of blindness in people around the world. Early diagnosis and treatment of DR can be accomplished by organizing large regular screening programs. Still, it is difficult to spot diabetic retinopathy timely because the situation might not indicate signs in the primary stages of the disease. Due to a drastic increase in diabetic patients, there is an urgent need for efficient diabetic retinopathy detecting systems. Auto-encoders, sparse coding, and limited Boltzmann machines were used as a few past deep learning (DL) techniques and features for the classification of DR. Convolutional Neural Networks (CNN) have been identified as a promising solution for detecting and classifying DR. We employ the deep learning capabilities of efficient net batch normalization (BNs) pre-trained models to automatically acquire discriminative features from fundus images. However, we successfully achieved F1 scores above 80% on all efficient net BNs in the EYE-PACS dataset (calculated F1 score for DeepDRiD another dataset) and the results are better than previous studies. In this paper, we improved the accuracy and F1 score of the efficient net BNs pre-trained models on the EYE-PACS dataset by applying a Gaussian Smooth filter and data augmentation transforms. Using our proposed technique, we have achieved F1 scores of 84% and 87% for EYE-PACS and DeepDRiD. [ABSTRACT FROM AUTHOR]

Electron energy loss spectroscopy (EELS) is a well established technique in electron microscopy that yields information on the elemental content of a sample in a very direct manner. One of the persisting limitations of EELS is the requirement for manual identification of core-loss edges and their corresponding elements. This can be especially bothersome in spectrum imaging, where a large amount of spectra are recorded when spatially scanning over a sample area. This paper introduces a synthetic dataset with 736,000 labeled EELS spectra, computed from available generalized oscillator strength tables, that represents 107 K, L, M or N core-loss edges and 80 chemical elements. Generic lifetime broadened peaks are used to mimic the fine structure due to band structure effects present in experimental core-loss edges. The proposed dataset is used to train and evaluate a series of neural network architectures, being a multilayer perceptron, a convolutional neural network, a U-Net, a residual neural network, a vision transformer and a compact convolutional transformer. An ensemble of neural networks is used to further increase performance. The ensemble network is used to demonstrate fully automated elemental mapping in a spectrum image, both by directly mapping the predicted elemental content and by using the predicted content as input for a physical model-based mapping. [ABSTRACT FROM AUTHOR]

An alarming number of fatalities caused by the COVID-19 pandemic has forced the scientific community to accelerate the process of therapeutic drug discovery. In this regard, the collaboration between biomedical scientists and experts in artificial intelligence (AI) has led to a number of in silico tools being developed for the initial screening of therapeutic molecules. All living organisms produce antiviral peptides (AVPs) as a part of their first line of defense against invading viruses. The Deep-AVPiden model proposed in this paper and its corresponding web app, deployed at https://deep-avpiden.anvil.app, is an effort toward discovering novel AVPs in proteomes of living organisms. Apart from Deep-AVPiden, a computationally efficient model called Deep-AVPiden (DS) has also been developed using the same underlying network but with point-wise separable convolutions. The Deep-AVPiden and Deep-AVPiden (DS) models show an accuracy of 90% and 88%, respectively, and both have a precision of 90%. Also, the proposed models were statistically compared using the Student's t-test. On comparing the proposed models with the state-of-the-art classifiers, it was found that they are much better than them. To test the proposed model, we identified some AVPs in the natural defense proteins of plants, mammals, and fishes and found them to have appreciable sequence similarity with some experimentally validated antimicrobial peptides. These AVPs can be chemically synthesized and tested for their antiviral activity. [ABSTRACT FROM AUTHOR]

Aphid infestation poses a significant threat to crop production, rural communities, and global food security. While chemical pest control is crucial for maximizing yields, applying chemicals across entire fields is both environmentally unsustainable and costly. Hence, precise localization and management of aphids are essential for targeted pesticide application. The paper primarily focuses on using deep learning models for detecting aphid clusters. We propose a novel approach for estimating infection levels by detecting aphid clusters. To facilitate this research, we have captured a large-scale dataset from sorghum fields, manually selected 5447 images containing aphids, and annotated each individual aphid cluster within these images. To facilitate the use of machine learning models, we further process the images by cropping them into patches, resulting in a labeled dataset comprising 151,380 image patches. Then, we implemented and compared the performance of four state-of-the-art object detection models (VFNet, GFLV2, PAA, and ATSS) on the aphid dataset. Extensive experimental results show that all models yield stable similar performance in terms of average precision and recall. We then propose to merge close neighboring clusters and remove tiny clusters caused by cropping, and the performance is further boosted by around 17%. The study demonstrates the feasibility of automatically detecting and managing insects using machine learning models. The labeled dataset will be made openly available to the research community. [ABSTRACT FROM AUTHOR]

Distributed learning, as the most popular solution for training large-scale data for deep learning, consists of multiple participants collaborating on data training tasks. However, the malicious behavior of some during the training process, like Byzantine participants who would interrupt or control the learning process, will trigger the crisis of data security. Although recent existing defense mechanisms use the variability of Byzantine node gradients to clear Byzantine values, it is still unable to identify and then clear the delicate disturbance/attack. To address this critical issue, we propose an algorithm named consensus aggregation in this paper. This algorithm allows computational nodes to use the information of verification nodes to verify the effectiveness of the gradient in the perturbation attack, reaching a consensus based on the effective verification of the gradient. Then the server node uses the gradient as the valid gradient for gradient aggregation calculation through the consensus reached by other computing nodes. On the MNIST and CIFAR10 datasets, when faced with Drift attacks, the proposed algorithm outperforms common existing aggregation algorithms (Krum, Trimmed Mean, Bulyan), with accuracies of 93.3%, 94.06% (MNIST dataset), and 48.66%, 51.55% (CIFAR10 dataset), respectively. This is an improvement of 3.0%, 3.8% (MNIST dataset), and 19.0%, 26.1% (CIFAR10 dataset) over the current state-of-the-art methods, and successfully defended against other attack methods. [ABSTRACT FROM AUTHOR]

Nowadays, global public health crises are occurring more frequently, and accurate prediction of these diseases can reduce the burden on the healthcare system. Taking COVID-19 as an example, accurate prediction of infection can assist experts in effectively allocating medical resources and diagnosing diseases. Currently, scholars worldwide use single model approaches or epidemiology models more often to predict the outbreak trend of COVID-19, resulting in poor prediction accuracy. Although a few studies have employed ensemble models, there is still room for improvement in their performance. In addition, there are only a few models that use the laboratory results of patients to predict COVID-19 infection. To address these issues, research efforts should focus on improving disease prediction performance and expanding the use of medical disease prediction models. In this paper, we propose an innovative deep learning model Whale Optimization Convolutional Neural Networks (CNN), Long-Short Term Memory (LSTM) and Artificial Neural Network (ANN) called WOCLSA which incorporates three models ANN, CNN and LSTM. The WOCLSA model utilizes the Whale Optimization Algorithm to optimize the neuron number, dropout and batch size parameters in the integrated model of ANN, CNN and LSTM, thereby finding the global optimal solution parameters. WOCLSA employs 18 patient indicators as predictors, and compares its results with three other ensemble deep learning models. All models were validated with train-test split approaches. We evaluate and compare our proposed model and other models using accuracy, F1 score, recall, AUC and precision metrics. Through many studies and tests, our results show that our prediction models can identify patients with COVID-19 infection at the AUC of 91%, 91%, and 93% respectively. Other prediction results achieve a respectable accuracy of 92.82%, 92.79%, and 91.66% respectively, f1-score of 93.41%, 92.79%, and 92.33% respectively, precision of 93.41%, 92.79%, and 92.33% respectively, recall of 93.41%, 92.79%, and 92.33% respectively. All of these exceed 91%, surpassing those of comparable models. The execution time of WOCLSA is also an advantage. Therefore, the WOCLSA ensemble model can be used to assist in verifying laboratory research results and predict and to judge various diseases in public health events. [ABSTRACT FROM AUTHOR]

In the field of machine intelligence and ubiquitous computing, there has been a growing interest in human activity recognition using wearable sensors. Over the past few decades, researchers have extensively explored learning-based methods to develop effective models for identifying human behaviors. Deep learning algorithms, known for their powerful feature extraction capabilities, have played a prominent role in this area. These algorithms can conveniently extract features that enable excellent recognition performance. However, many successful deep learning approaches have been built upon complex models with multiple hyperparameters. This paper examines the current research on human activity recognition using deep learning techniques and discusses appropriate recognition strategies. Initially, we employed multiple convolutional neural networks to determine an effective architecture for human activity recognition. Subsequently, we developed a hybrid convolutional neural network that incorporates a channel attention mechanism. This mechanism enables the network to capture deep spatio-temporal characteristics in a hierarchical manner and distinguish between different human movements in everyday life. Our investigations, using the UCI-HAR, WISDM, and IM-WSHA datasets, demonstrated that our proposed model, which includes cross-channel multi-size convolution transformations, outperformed previous deep learning architectures with accuracy rates of 98.92%, 98.80%, and 98.45% respectively. These results indicate that the suggested model surpasses state-of-the-art approaches in terms of overall accuracy, as supported by the research findings. [ABSTRACT FROM AUTHOR]

Exposure to alcohol content in media increases alcohol consumption and related harm. With exponential growth of media content, it is important to use algorithms to automatically detect and quantify alcohol exposure. Foundation models such as Contrastive Language-Image Pretraining (CLIP) can detect alcohol exposure through Zero-Shot Learning (ZSL) without any additional training. In this paper, we evaluated the ZSL performance of CLIP against a supervised algorithm called Alcoholic Beverage Identification Deep Learning Algorithm Version-2 (ABIDLA2), which is specifically trained to recognise alcoholic beverages in images, across three tasks. We found ZSL achieved similar performance compared to ABIDLA2 in two out of three tasks. However, ABIDLA2 outperformed ZSL in a fine-grained classification task in which determining subtle differences among alcoholic beverages (including containers) are essential. We also found that phrase engineering is essential for improving the performance of ZSL. To conclude, like ABIDLA2, ZSL with little phrase engineering can achieve promising performance in identifying alcohol exposure in images. This makes it easier for researchers, with little or no programming background, to implement ZSL effectively to obtain insightful analytics from digital media. Such analytics can assist researchers and policy makers to propose regulations that can prevent alcohol exposure and eventually prevent alcohol consumption. [ABSTRACT FROM AUTHOR]

This paper presents a new strategy to generate synthetic samples containing casting defects. Four samples of Inconel 100 containing casting defects such as shrinkages and pores have been characterized using X-ray tomography and are used as reference for this application. Shrinkages are known to be tortuous in shape and more detrimental for the mechanical properties of materials, especially metal fatigue, whereas pores can be of two types: broken shrinkage pores with arbitrary shape and gaseous pores of spherical shape. For the generation of synthetic samples, an integrated module of Spatial Point Pattern (SPP) analysis and deep learning techniques such as Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNNs) are used. The SPP analysis describes the spatial distributions of casting defects in material space, whereas GANs and CNNs generate a defect of arbitrary morphology very close to real defects. SPP analysis reveals the existence of two different void nucleation mechanisms during metal solidification associated to shrinkages and pores. Our deep learning model successfully generates casting defects with defect size ranging from 100 µm to 1.5 mm and of very realistic shapes. The entire synthetic microstructure generation process respects the global defect statistics of reference samples and the generated samples are validated by statistically comparing with real samples. [ABSTRACT FROM AUTHOR]

Federated learning enables multiple nodes to perform local computations and collaborate to complete machine learning tasks without centralizing private data of nodes. However, the frequent model gradients upload/download operations required by the framework result in high communication costs, which have become the main bottleneck for federated learning as deep models scale up, hindering its performance. In this paper, we propose a two-layer accumulated quantized compression algorithm (TLAQC) that effectively reduces the communication cost of federated learning. TLAQC achieves this by reducing both the cost of individual communication and the number of global communication rounds. TLAQC introduces a revised quantization method called RQSGD, which employs zero-value correction to mitigate ineffective quantization phenomena and minimize average quantization errors. Additionally, TLAQC reduces the frequency of gradient information uploads through an adaptive threshold and parameter self-inspection mechanism, further reducing communication costs. It also accumulates quantization errors and retained weight deltas to compensate for gradient knowledge loss. Through quantization correction and two-layer accumulation, TLAQC significantly reduces precision loss caused by communication compression. Experimental results demonstrate that RQSGD achieves an incidence of ineffective quantization as low as 0.003% and reduces the average quantization error to 1.6 × 10 - 5 . Compared to full-precision FedAVG, TLAQC compresses uploaded traffic to only 6.73% while increasing accuracy by 1.25%. [ABSTRACT FROM AUTHOR]

Deep learning (DL) based detection models are powerful tools for large-scale analysis of dynamic biological behaviors in video data. Supervised training of a DL detection model often requires a large amount of manually-labeled training data which are time-consuming and labor-intensive to acquire. In this paper, we propose LFAGPA (Learn From Algorithm-Generated Pseudo-Annotations) that utilizes (noisy) annotations which are automatically generated by algorithms to train DL models for ant detection in videos. Our method consists of two main steps: (1) generate foreground objects using a (set of) state-of-the-art foreground extraction algorithm(s); (2) treat the results from step (1) as pseudo-annotations and use them to train deep neural networks for ant detection. We tackle several challenges on how to make use of automatically generated noisy annotations, how to learn from multiple annotation resources, and how to combine algorithm-generated annotations with human-labeled annotations (when available) for this learning framework. In experiments, we evaluate our method using 82 videos (totally 20,348 image frames) captured under natural conditions in a tropical rain-forest for dynamic ant behavior study. Without any manual annotation cost but only algorithm-generated annotations, our method can achieve a decent detection performance (77% in F 1 score). Moreover, when using only 10% manual annotations, our method can train a DL model to perform as well as using the full human annotations (81% in F 1 score). [ABSTRACT FROM AUTHOR]

Ultrasonic vocalizations (USVs) analysis represents a fundamental tool to study animal communication. It can be used to perform a behavioral investigation of mice for ethological studies and in the field of neuroscience and neuropharmacology. The USVs are usually recorded with a microphone sensitive to ultrasound frequencies and then processed by specific software, which help the operator to identify and characterize different families of calls. Recently, many automated systems have been proposed for automatically performing both the detection and the classification of the USVs. Of course, the USV segmentation represents the crucial step for the general framework, since the quality of the call processing strictly depends on how accurately the call itself has been previously detected. In this paper, we investigate the performance of three supervised deep learning methods for automated USV segmentation: an Auto-Encoder Neural Network (AE), a U-NET Neural Network (UNET) and a Recurrent Neural Network (RNN). The proposed models receive as input the spectrogram associated with the recorded audio track and return as output the regions in which the USV calls have been detected. To evaluate the performance of the models, we have built a dataset by recording several audio tracks and manually segmenting the corresponding USV spectrograms generated with the Avisoft software, producing in this way the ground-truth (GT) used for training. All three proposed architectures demonstrated precision and recall scores exceeding 90 % , with UNET and AE achieving values above 95 % , surpassing other state-of-the-art methods that were considered for comparison in this study. Additionally, the evaluation was extended to an external dataset, where UNET once again exhibited the highest performance. We suggest that our experimental results may represent a valuable benchmark for future works. [ABSTRACT FROM AUTHOR]

Diabetic Retinopathy (DR) is a major cause of blindness worldwide. Early detection and treatment are crucial to prevent vision loss, making accurate and timely diagnosis critical. Deep learning technology has shown promise in the automated diagnosis of DR, and in particular, multi-lesion segmentation tasks. In this paper, we propose a novel Transformer-based model for DR segmentation that incorporates hyperbolic embeddings and a spatial prior module. The proposed model is primarily built on a traditional Vision Transformer encoder and further enhanced by incorporating a spatial prior module for image convolution and feature continuity, followed by feature interaction processing using the spatial feature injector and extractor. Hyperbolic embeddings are used to classify feature matrices from the model at the pixel level. We evaluated the proposed model's performance on the publicly available datasets and compared it with other widely used DR segmentation models. The results show that our model outperforms these widely used DR segmentation models. The incorporation of hyperbolic embeddings and a spatial prior module into the Vision Transformer-based model significantly improves the accuracy of DR segmentation. The hyperbolic embeddings enable us to better capture the underlying geometric structure of the feature matrices, which is important for accurate segmentation. The spatial prior module improves the continuity of the features and helps to better distinguish between lesions and normal tissues. Overall, our proposed model has potential for clinical use in automated DR diagnosis, improving accuracy and speed of diagnosis. Our study shows that the integration of hyperbolic embeddings and a spatial prior module with a Vision Transformer-based model improves the performance of DR segmentation models. Future research can explore the application of our model to other medical imaging tasks, as well as further optimization and validation in real-world clinical settings. [ABSTRACT FROM AUTHOR]

Deep learning is currently being used to automate surface defect detection in aluminum. The common target detection models based on neural networks often have a large number of parameters and a slow detection speed, which is not conducive to real-time detection. Therefore, this paper proposes a lightweight aluminum surface defect detection model, M2-BL-YOLOv4, based on the YOLOv4 algorithm. First, in the YOLOv4 model, the complex CSPDarkNet53 backbone network was modified into an inverted residual structure, which greatly reduced the number of parameters in the model and increased the detection speed. Second, a new feature fusion network, BiFPN-Lite, is designed to improve the fusion ability of the network and further improve its detection accuracy. The final results show that the mean average precision of the improved lightweight YOLOv4 algorithm in the aluminum surface defect test set reaches 93.5%, the number of model parameters is reduced to 60% of the original, and the number of frames per second (FPS) detected is 52.99, which increases the detection speed by 30%. The efficient detection of aluminum surface defects is realized. [ABSTRACT FROM AUTHOR]

Automatic modulation recognition (AMR) is a critical technology in spatial cognitive radio (SCR), and building high-performance AMR model can achieve high classification accuracy of signals. AMR is a classification problem essentially, and deep learning has achieved excellent performance in various classification tasks. In recent years, joint recognition of multiple networks has become increasingly popular. In complex wireless environments, there are multiple signal types and diversity of characteristics between different signals. Also, the existence of multiple interference in wireless environment makes the signal characteristics more complex. It is difficult for a single network to accurately extract the unique features of all signals and achieve accurate classification. So, this article proposes a time–frequency domain joint recognition model that combines two deep learning networks (DLNs), to achieve higher accuracy AMR. A DLN named MCLDNN (multi-channel convolutional long short-term deep neural network) is trained on samples composed of in-phase and quadrature component (IQ) signals, to distinguish modulation modes that are relatively easy to identify. This paper proposes a BiGRU3 (three-layer bidirectional gated recurrent unit) network based on FFT as the second DLN. For signals with significant similarity in the time domain and significant differences in the frequency domain that are difficult to distinguish by the former DLN, such as AM-DSB and WBFM, FFT (Fast Fourier Transform) is used to obtain frequency domain amplitude and phase (FDAP) information. Experiments have shown that the BiGUR3 network has superior extraction performance for amplitude spectrum and phase spectrum features. Experiments are conducted on two publicly available datasets, the RML2016.10a and RML2016.10b, and the results show that the overall recognition accuracy of the proposed joint model reaches 94.94% and 96.69%, respectively. Compared to a single network, the recognition accuracy is significantly improved. At the same time, the recognition accuracy of AM-DSB and WBFM signals has been improved by 17% and 18.2%, respectively. [ABSTRACT FROM AUTHOR]

As an effective approach to perceive environments, acoustic scene classification (ASC) has received considerable attention in the past few years. Generally, ASC is deemed a challenging task due to subtle differences between various classes of environmental sounds. In this paper, we propose a novel approach to perform accurate classification based on the aggregation of spatial–temporal features extracted from a multi-branch three-dimensional (3D) convolution neural network (CNN) model. The novelties of this paper are as follows. First, we form multiple frequency-domain representations of signals by fully utilizing expert knowledge on acoustics and discrete wavelet transformations (DWT). Secondly, we propose a novel 3D CNN architecture featuring residual connections and squeeze-and-excitation attentions (3D-SE-ResNet) to effectively capture both long-term and short-term correlations inherent in environmental sounds. Thirdly, an auxiliary supervised branch based on the chromatogram of the original signal is incorporated in the proposed architecture to alleviate overfitting risks by providing supplementary information to the model. The performance of the proposed multi-input multi-feature 3D-CNN architecture is numerically evaluated on a typical large-scale dataset in the 2019 IEEE AASP Challenge on Detection and Classification of Acoustic Scenes and Events (DCASE 2019) and is shown to obtain noticeable performance gains over the state-of-the-art methods in the literature. [ABSTRACT FROM AUTHOR]

Accurate and efficient real-time detection of orchard pests was essential and could improve the economic benefits of the fruit industry. The orchard pest dataset, PestImgData, was built through a series of methods such as web crawler, specimen image collection and data augmentation. PestImgData was composed of two parts, PestImgData-1 and PestImgData-2. It contained 24,796 color images and covered 7 types of orchard pests. Based on the PestImgData and YOLOv4 algorithm, this paper conducted a preliminary study on the real-time object detection of orchard pests from 4 perspectives: transfer learning, activation function, anchor box, and batch normalization. In addition, this paper also visualized the feature learning ability of the detection models. On the basis of the above research, three improvement measures were adopted: the post-processing NMS algorithm was upgraded to DIoU-NMS, the training method was upgraded to 2-time finetuning training and the training data was enhanced. The performance of the improved model, F-D-YOLOv4-PEST, had been effectively improved. The mean average precision of F-D-YOLOv4-PEST was 92.86%, and the detection time of a single picture was 12.22 ms, which could meet the real-time detection requirements. In addition, in the case of high overlap area or high density, F-D-YOLOv4-PEST still maintained good performance. In the testing process of the laboratory and the greenhouse, including the wired network and the wireless network, F-D-YOLOv4-PEST could locate and classify pests as expected. This research could provide technical reference for the intelligent identification of agricultural pests based on deep learning. [ABSTRACT FROM AUTHOR]

Wetland vegetation classification using deep learning algorithm and unmanned aerial vehicle (UAV) images have attracted increased attentions. However, there exist several challenges in mapping karst wetland vegetation due to its fragmentation, intersection, and high heterogeneity of vegetation patches. This study proposed a novel approach to classify karst vegetation in Huixian National Wetland Park, the largest karst wetland in China by fusing single-class SegNet classification using the maximum probability algorithm. A new optimized post-classification algorithm was developed to eliminate the stitching traces caused by SegNet model prediction. This paper evaluated the effect of multi-class and fusion of multiple single-class SegNet models with different EPOCH values on mapping karst vegetation using UAV images. Finally, this paper carried out a comparison of classification accuracies between object-based Random Forest (RF) and fusion of single-class SegNet models. The specific conclusions of this paper include the followings: (1) fusion of four single-class SegNet models produced better classification for karst wetland vegetation than multi-class SegNet model, and achieved the highest overall accuracy of 87.34%; (2) the optimized post-classification algorithm improved classification accuracy of SegNet model by eliminating splicing traces; (3) classification performance of single-class SegNet model outperformed multi-class SegNet model, and improved classification accuracy (F1-Score) ranging from 10 to 25%; (4) Fusion of single-class SegNet models and object-based RF classifier both produced good classifications for karst wetland vegetation, and achieved over 87% overall accuracy. [ABSTRACT FROM AUTHOR]