Search

Over the last two decades, research in hyperspectral imaging has been increasing and its use in horticulture is expected to be spreading in the coming years. The emerging techniques are currently gaining interest of the research community. However, there are still challenges to the applicability. In this review we demonstrate that hyperspectral imaging can be used as an effective tool for fruit, vegetables and mushrooms in assessing quality parameters related to well defined variables that can be analysed in the laboratory, as well as complex properties such as maturity, ripeness, detection of biotic defects, physiological disorders, mechanical damages, and sensory quality. Therefore, this paper starts by giving an overview of the quality concept of produce, measuring principles, theory and analysis of hyperspectral imaging systems. Then, emerging techniques to monitor and assess quality parameters, both pre- and postharvest, are described, as well as applications of these are reviewed and discussed. Afterwards, this review proceeds by illustrating the current and potential use of artificial intelligence and its subdomains, machine learning and deep learning, for hyperspectral imaging analysis in horticulture. Lastly, some challenges and considerations for future research are highlighted, including improvement of data availability, possible solutions for an improved integration of artificial intelligence and the transfer of knowledge from research parameters to parameters relevant for industrial stakeholders. • Hyperspectral imaging is an effective tool for in assessing quality parameters. • The most abundantly used wavelengths are 601–850 nm, used in over 50% of studies. • PLSR and SVM are most commonly used models for machine learning. • Deep learning in combination with hyperspectral data is unexplored in horticulture. • Improving availability, AI knowledge transfer and standardisation is recommended. [ABSTRACT FROM AUTHOR]

• In-depth review of deep learning methods for seizure video analysis (last 7 years). • Introducing a multi-stream framework to promote AI adoption in Epilepsy. • Unveiling solutions in AI for seizure analysis challenges overlooked in prior reviews. Seizure events can manifest as transient disruptions in the control of movements which may be organized in distinct behavioral sequences, accompanied or not by other observable features such as altered facial expressions. The analysis of these clinical signs, referred to as semiology, is subject to observer variations when specialists evaluate video-recorded events in the clinical setting. To enhance the accuracy and consistency of evaluations, computer-aided video analysis of seizures has emerged as a natural avenue. In the field of medical applications, deep learning and computer vision approaches have driven substantial advancements. Historically, these approaches have been used for disease detection, classification, and prediction using diagnostic data; however, there has been limited exploration of their application in evaluating video-based motion detection in the clinical epileptology setting. While vision-based technologies do not aim to replace clinical expertise, they can significantly contribute to medical decision-making and patient care by providing quantitative evidence and decision support. Behavior monitoring tools offer several advantages such as providing objective information, detecting challenging-to-observe events, reducing documentation efforts, and extending assessment capabilities to areas with limited expertise. The main applications of these could be (1) improved seizure detection methods; (2) refined semiology analysis for predicting seizure type and cerebral localization. In this paper, we detail the foundation technologies used in vision-based systems in the analysis of seizure videos, highlighting their success in semiology detection and analysis, focusing on work published in the last 7 years. We systematically present these methods and indicate how the adoption of deep learning for the analysis of video recordings of seizures could be approached. Additionally, we illustrate how existing technologies can be interconnected through an integrated system for video-based semiology analysis. Each module can be customized and improved by adapting more accurate and robust deep learning approaches as these evolve. Finally, we discuss challenges and research directions for future studies. [ABSTRACT FROM AUTHOR]

• A new adaptive blind image watermarking technology, utilizing artificial intelligence (AI) and named AwDD, has been proposed for detecting color image deepfakes. • The AI technology used includes face detection, denoising autoencoder (DAE), and the gray wolf optimizer (GWO). • The proposed method employs mixed modulation (MM) with partly sign-altered mean value (PSAMV) to embed a set of coefficients, thereby increasing robustness against attacks while maintaining high image quality. • Blind adaptive deepfake detection (BADD) with the tamper detection mean value (TDMV) is applied to adaptively detect changes in the relative position when a face image is modified or deepfaked. • The GWO optimizes parameters to enhance imperceptibility and robustness, while the DAE improves the visual recognizability of the extracted watermark. This paper proposes a new adaptive blind watermarking technology for deepfake detection, which can embed deepfake detection information into the image and verify the image's authenticity without requiring additional information. The proposed scheme utilizes mixed modulation combined with partly sign-altered mean value to embed a set of coefficients that enhance robustness against attacks while maintaining high image quality. Additionally, blind adaptive deepfake detection technology with the tamper detection mean value is employed to detect relative positions adaptively, even when face images are slightly modified or deepfaked. To further improve the performance of the proposed scheme, a gray wolf optimizer is introduced to optimize parameters, and a denoising autoencoder is employed to facilitate the identification of extracted watermarks. This technology will adaptively embed watermark information while preserving the original face image, thereby maintaining the authenticity of the face in the image and verifying the owner of the image. The code is available at https://github.com/lyhsu01/AwDD. [ABSTRACT FROM AUTHOR]

Utilizing the latest advances in Artificial Intelligence (AI), the computer vision community is now witnessing an unprecedented evolution in all kinds of perception tasks, particularly in object detection. Based on multiple spatially separated perception nodes, Cooperative Perception (CP) has emerged to significantly advance the perception of automated driving. However, current cooperative object detection methods mainly focus on ego-vehicle efficiency without considering the practical issues of system-wide costs. In this paper, we introduce VINet, a unified deep learning-based CP network for scalable, lightweight, and heterogeneous cooperative 3D object detection. VINet is the first CP method designed from the standpoint of large-scale system-level implementation and can be divided into three main phases: (1) Global Pre-Processing and Lightweight Feature Extraction which prepare the data into global style and extract features for cooperation in a lightweight manner; (2) Two-Stream Fusion which fuses the features from scalable and heterogeneous perception nodes; and (3) Central Feature Backbone and 3D Detection Head which further process the fused features and generate cooperative detection results. An open-source data experimental platform is designed and developed for CP dataset acquisition and model evaluation. The experimental analysis shows that VINet can reduce 84% system-level computational cost and 94% system-level communication cost while improving the 3D detection accuracy. [ABSTRACT FROM AUTHOR]

• implement all-optical neural networks by combining the techniques of time-stretching and time-division multiplexing. • enabling classification of different datasets. • this architecture is fast, operates efficiently, and achieves pleasing accuracy rates. As factors such as power consumption and bandwidth constrain the development of electronic computing in deep learning, optical neuromorphic computing has the potential to meet the growing computational demands of artificial intelligence, presenting various opportunities and challenges. In this paper, we propose an all-optical neural network through a combination of time-stretching and time-division multiplexing that can be used for the classification of multiple datasets. The system modulates a large number of parameters into chirp optical pulses by time-stretching and increases operational efficiency in conjunction with time-division multiplexing. This all-optical neural network system proposed is capable of operating at the speed of 1 Tera-OPS/s and can achieve data transfer rate in excess of 800 Gbit/s. The system is tested by the feedforward neural network, and the computational simulation realized very encouraging and considerable results. [ABSTRACT FROM AUTHOR]

Neural networks (NN) have spurred significant interest in automating machine condition monitoring, with many recent studies focusing on networks that take raw diagnostic signals as input. However, the complexity of these NNs has raised concerns about their physical justification and trustworthiness. Although explainable artificial intelligence (XAI) has developed promising tools to address these issues, the nature and size of raw diagnostic inputs have limited the effectiveness of XAI techniques in condition monitoring. This paper proposes a novel approach combining domain transformation and discretisation techniques with the XAI technique of Shapely value additive explanation (SHAP). This methodology allows the input domain for the network to remain in the original time-domain, while the explanation domain can be chosen as frequency or time–frequency and its discretisation adapted to the specific diagnostic case. This flexibility makes networks trained on raw signals more interpretable, while the controllable domain discretisation facilitates the implementation of XAI with manageable computational costs. The proposed methodology is tested on numerical cases and experimental signals, including the UNSW gear-wear and CWRU bearing-fault datasets. This enhancement of XAI is applicable to common black-box classifiers, providing researchers with an effective tool to give physical justification to their models and bolster trust in AI for machine condition monitoring applications. • Addressing interpretability in time-domain processing NNs for condition monitoring. • Novel methodology to explain a NN in a transformed domain to their processed data. • Novel discretisation techniques to decrease computational costs of the calculation. • Application to two numerical examples and gear and bearing fault diagnostics. [ABSTRACT FROM AUTHOR]

Feature pyramid network (FPN) is a typical detector commonly for solving the issue of object detection at different scales. However, the lateral connections in FPN lead to the loss of feature information due to the reduction of feature channels. Moreover, the top-down feature fusion will weaken the feature representation in the process of feature delivery because of features with different semantic information. In this paper, we propose a feature pyramid network with channel and content adaptive feature enhancement module (CCA-FPN), which uses a channel adaptive guided mechanism module (CAGM) and multi-scale content adaptive feature enhancement module (MCAFEM) to alleviate these problems. We conduct comprehensive experiments on the MS COCO dataset. By replacing FPN with CCA-FPN in ATSS, our models achieve 1.3 percentage points higher Average Precision (AP) when using ResNet50 as backbone. Furthermore, our CCA-FPN achieves 0.3 percentage points higher than the AugFPN which is the state-of-the-art FPN-based detector. • Targets of different scales often use feature pyramid networks for hierarchical detection. • Using deep learning for object detection, feature enhancement is beneficial for improving detection performance. • When performing feature fusion on the feature layer, it is necessary to reduce the channel dimension. • The reduction of the channel dimension can use the attention mechanism to mine the correlation between channels for compression. • In the learning process of neural networks, the learning target features are adapted using different weight sizes. [ABSTRACT FROM AUTHOR]

The detection and localization of damage in metallic structures using acoustic emission (AE) monitoring and artificial intelligence technology such as deep learning has been widely studied. However, a current challenge of this approach is the difficulty of obtaining sufficient labeled historical AE signals for the training process of deep learning models. This problem can be approached through the implementation of transfer learning. The innovation of this paper lies in the development of a transfer learning approach for AE source localization on a stainless-steel structure when no historical labeled AE signals are available for training. A finite element model is developed to generate numerical AE signals for the training. Unsupervised domain adaptation (UDA) technology is utilized to reduce the distribution difference between the numerical and the realistic AE signals and to derive the localization results of the unlabeled realistic AE signals. The results suggest that the proposed approach is capable of localizing AE signals with high accuracy in the absence of labeled training data. [ABSTRACT FROM AUTHOR]

Neural network pruning techniques can be effective in accelerating neural network models, making it possible to deploy them on edge devices. In this paper, we propose to prune neural networks using data variance. Unlike other existing methods, this method is somewhat robust and does not invalidate our criteria depending on the number of data batches and the number of training sessions. We also propose a pruning compensation technique. This technique fuses the pruned convolutional information into the remaining convolutional kernel close to it. This fusion operation can effectively help retain the pruned information. We evaluate the proposed method on a number of standard datasets and compare it with several current state-of-the-art methods. Our method always achieves better performance. For example, on Tiny ImageNet, our method can prune 54.2% FLOPs of ResNet50 while obtaining a 0.22% accuracy improvement. • A data-driven network pruning method with high robustness is proposed. • The proposed data-driven pruning method consumes less resources and is highly efficient. • A pruning compensation technique is proposed to reduce the negative impact of pruning. • The effectiveness of the proposed method is validated on several datasets. [ABSTRACT FROM AUTHOR]

[Display omitted] • Providing an overview of the cross disciplinary knowledge to building models in silico. • Review of the challenges in machine learning-based de novo model building methods. • Review of the machine learning modeling methods we have developed in the past decade. • Discussion on the current bottlenecks in de novo methods and future directions. The study of macromolecular structures has expanded our understanding of the amazing cell machinery and such knowledge has changed how the pharmaceutical industry develops new vaccines in recent years. Traditionally, X-ray crystallography has been the main method for structure determination, however, cryogenic electron microscopy (cryo-EM) has increasingly become more popular due to recent advancements in hardware and software. The number of cryo-EM maps deposited in the EMDataResource (formerly EMDatabase) since 2002 has been dramatically increasing and it continues to do so. De novo macromolecular complex modeling is a labor-intensive process, therefore, it is highly desirable to develop software that can automate this process. Here we discuss our automated, data-driven, and artificial intelligence approaches including map processing, feature extraction, modeling building, and target identification. Recently, we have enabled DNA/RNA modeling in our deep learning-based prediction tool, DeepTracer. We have also developed DeepTracer-ID, a tool that can identify proteins solely based on the cryo-EM map. In this paper, we will present our accumulated experiences in developing deep learning-based methods surrounding macromolecule modeling applications. [ABSTRACT FROM AUTHOR]

Nowadays, there is an increasing use of digital technologies and Artificial Intelligence (AI) such as Machine Learning (ML) models that leverage data to optimize the performances of systems in almost all activity sectors, including ML models for optimizing solutions related to CO2 capture from the atmosphere or CO2 emissions reduction from human activities. However, on the other hand, the use of AI models is leading to an increasing energy consumption that also raises environmental issues (in terms of CO2 emissions) which are less studied in the literature. This then raises the new question of a more realistic estimate of the carbon footprint (CO2 emissions in particular) of AI models in general, and particularly AI models aimed at reducing CO2 emissions. Thus, in this paper, for an AI model in this latter context, we propose a method to quantify both his negative impacts (quantity of CO2 emissions emitted by the training and use of the model) and his positive impacts (quantity of CO2 emissions saved when the model is used). The method is evaluated with three state-of-the-art AI models: (i) an artificial neural network model for managing the energy demand of Brazilian households, (ii) an adaptive neuro-fuzzy inference system for photovoltaic power forecast in Tunisia, (iii) and a Bayesian regression model for the electric vehicle routing problem in Sweden and Luxembourg. Results show that, if only the positive impacts are considered, the reduction of CO2 emitted due to the usage of the models is significant, but depends on each context (34%, 73%, and 9%, respectively). However, when both positive and negative impacts are considered, the negative impacts are sometimes higher than the positive impacts (the first and the third model) for a nominal use (1 user). Nevertheless, the balance becomes highly positive again, when these two projects are scaled up (realistic projections with many users). The second model cannot be scaled up, but the balance remains positive, even if the gains are much smaller. More generally, the CO2 emissions gain metrics provided by our method can be used as new metrics for comparing the efficiency of AI models (for reducing CO2 emissions) beyond predictive capacity-based traditional ML evaluation metrics. Based on the lessons learned from our study, we also provide seven global recommendations that can contribute to the reduction of the carbon footprint of ML models in general. • Increase of the use of AI models for optimizing solutions to reduce CO2 emissions. • Training and using AI models also generates CO2 emissions. • Method to quantify positive and negative environmental impacts of AI models. • Evaluations based on three state-of-the-art AI models. • Seven global recommendations for reducing CO2 emissions of AI models. [ABSTRACT FROM AUTHOR]

To solve the problem of low sign language recognition rate under the condition of small samples, a simple and effective static gesture recognition method based on an attention mechanism is proposed. The method proposed in this paper can enhance the features of both the details and the subject of the gesture image. The input of the proposed method depends on the intermediate feature map generated by the original network. Also, the proposed convolutional model is a lightweight general module, which can be seamlessly integrated into any CNN(Convolutional Neural Network) architecture and achieve significant performance gains with minimal overhead. Experiments on two different datasets show that the proposed method is effective and can improve the accuracy of sign language recognition of the benchmark model, making its performance better than the existing methods. [ABSTRACT FROM AUTHOR]

Liver cancers are the fourth leading cause of cancer-related mortality worldwide. In the past decade, breakthroughs in the field of artificial intelligence (AI) have inspired development of algorithms in the cancer setting. A growing body of recent studies have evaluated machine learning (ML) and deep learning (DL) algorithms for pre-screening, diagnosis and management of liver cancer patients through diagnostic image analysis, biomarker discovery and predicting personalized clinical outcomes. Despite the promise of these early AI tools, there is a significant need to explain the 'black box' of AI and work towards deployment to enable ultimate clinical translatability. Certain emerging fields such as RNA nanomedicine for targeted liver cancer therapy may also benefit from application of AI, specifically in nano-formulation research and development given that they are still largely reliant on lengthy trial-and-error experiments. In this paper, we put forward the current landscape of AI in liver cancers along with the challenges of AI in liver cancer diagnosis and management. Finally, we have discussed the future perspectives of AI application in liver cancer and how a multidisciplinary approach using AI in nanomedicine could accelerate the transition of personalized liver cancer medicine from bench side to the clinic. [Display omitted] [ABSTRACT FROM AUTHOR]

• This study provides new insights into the evaluation of automobile sound quality from the perspective of physiological acoustics and artificial intelligence. • An EEG test paradigm designed here provides the experimental guidance for evaluating specific sound quality with physiological signals; • The advantages of CNN in spatial characteristics and the LSTM in temporal characteristics are combined here to construct an HDNN to adaptively extract EEG features, which is 14.6% more accurate than the manual extraction of EEG feature. • The score prediction of sportiness sound quality fused with EEG signals is achieved by using the HDNN, which outperforms the traditional CNN and LSTM. The evaluation of automobile sound quality is not only related to the inherent properties of the sound, but also to the psychological and physiological state of the evaluator, which makes the evaluation of sound quality become an interdisciplinary research. However, there are some deficiencies regarding the existing studies as follows: (1) it is imprecise enough to visualize the preference of evaluators for sound on the scale; (2) the evaluators without acoustic experience are prone to deviations in their evaluations, resulting in the universally and unapplicable results; (3) the automobile sound quality cannot be fully reflected only by physical and psychological parameters, and most of the existing evaluation models of sound quality are only characterized by shallow architectures. To alleviate the above flaws, a hybrid deep neural network (HDNN) is constructed to achieve the evaluation of automobile interior acceleration sound fused with physiological signals in this paper. An EEG test paradigm under the stimulation of automobile sound is meanwhile designed for a feasibility confirmation of the proposed method. In addition, the acquired EEG datum are respectively analyzed from two perspectives: the manual extraction of typical EEG features and the adaptive extraction of EEG features based on HDNN model. Furthermore, the performance of the proposed HDNN is also validated by comparing with conventional convolutional neural network (CNN) and long short time memory (LSTM). The results indicate that the adaptively extracted EEG features are superior to the manually extracted EEG features, and the proposed HDNN outperforms the CNN and LSTM in terms of accuracy, sensitivity as well as precision, and the optimal accuracy of HDNN is up to 88.1%. Thus, the constructed HDNN contributes to achieving accurate evaluation of automobile sound quality with EEG signal. [ABSTRACT FROM AUTHOR]

With the great success of deep neural networks, combining deep learning with traditional dictionary learning has become a hot issue. However, the performance of these methods is still limited for several reasons. First, some existing methods update dictionary learning and classifier as two independent modules, which limits the classification performance. Second, the non-attention dictionary is learned to represent all images, reducing the model representation flexibility. In this paper, we design a novel end-to-end model named Multi-layer Attention Dictionary Pair Learning Network (MADPL-net), which integrates the learning schemes of the convolutional neural network, deep encoder learning, and attention dictionary pair learning (ADicL) into a unified framework. The encoder layer contains the ADicL block, which selects more image-attentive atoms in the dictionary pair block via the softmax function to ensure MADPL-net classification capability. In addition, ADicL schema can yield discriminative dictionary atoms and feature maps with high inter-class separation and high intra-class compactness. To improve the sparse representation learning performance, MADPL-net adds l 1 − norm constraint of the analysis dictionary to the cross-entropy loss function. Extensive experiments show that MADPL-net can achieve excellent performance over other state-of-the-arts. • A powerful architecture, called the multi-layer attention dictionary pair learning network (MADPL-net), is proposed for image classification. • The MADPL-net integrates convolutional neural network, deep encoder learning, and attention dictionary pair learning into a unified framework. • The MADPL-net applies the attentional dictionary learning block to select more image-attentive atoms via the soft-max function. • The MADPL-net adds the l 1 − norm constraint of analysis dictionary to the crossentropy loss function. • The MADPL-net can ensure discriminative dictionary atoms and feature maps with high inter-class separation and high intra-class compactness. [ABSTRACT FROM AUTHOR]

Rapid advances in artificial intelligence (AI) in the last decade have been largely built upon the wide applications of deep learning (DL). However, the high carbon footprint yielded by larger and larger DL networks has become a concern for sustainability. Furthermore, DL decision mechanism is somewhat obscure in that it can only be verified by test data. Green learning (GL) is being proposed as an alternative paradigm to address these concerns. GL is characterized by low carbon footprints, lightweight model, low computational complexity, and logical transparency. It offers energy-efficient solutions in cloud centers as well as mobile/edge devices. GL also provides a more transparent, logical decision-making process which is essential to gaining people's trust. Several statistical tools such as unsupervised representation learning, supervised feature learning, and supervised decision learning, have been developed to achieve this goal in recent years. We have seen a few successful GL examples with performance comparable with state-of-the-art DL solutions. This paper introduces the key characteristics of GL, its demonstrated applications, and future outlook. [ABSTRACT FROM AUTHOR]

Although the Arabic language is the fourth most popular language in the world, it has not received sufficient attention in artificial intelligence research, especially in automatic speech recognition (ASR). The key feature of the Arabic language is that its words are pronounced exactly as they are written. Above all, taking into account the diacritics, 2 2 Throughout this paper, the alphabet is considered with diacritics. there are no words with similar pronunciation and writing. This motivates us to think of building an Arabic ASR system by recognizing its alphabet phonetics. Therefore, the Arabic alphabet phonemes classification must be studied, this is what the paper aims to achieve. In this paper, we create a new dataset, called Arabic alphabet phonetics dataset (AAPD). AAPD was collected by taking sound recordings of 1420 persons. We build several Arabic alphabet phonemes classification systems using three feature extraction techniques and four deep neural networks. Based on AAPD, we designed numerous experiments to compare the performance of feature extraction and classification methods, which can be used as a benchmark. Experimental results showed that Mel-frequency Cepstral Coefficient (MFCC) is considered most effective to feature extraction due to its highest accuracy, particularly when using 20 for Mel-bands number the training time is the least. Additionally, the appropriate model that achieved the highest accuracy with the least computational load is the proposed model VGG–based, where acquired an accuracy of 95.68%. • Create a novel phonemes dataset of the Arabic alphabet with diacritics called AAPD. • Evaluate three popular techniques' performance for features extraction. • Provide a classification benchmark for the AAPD dataset using deep Neural Networks. [ABSTRACT FROM AUTHOR]

Methods for linking individuals across historical data sets, typically in combination with AI based transcription models, are developing rapidly. Perhaps the single most important identifier for linking is personal names. However, personal names are prone to enumeration and transcription errors and although modern linking methods are designed to handle such challenges, these sources of errors are critical and should be minimized. For this purpose, improved transcription methods and large-scale databases are crucial components. This paper describes and provides documentation for HANA, a newly constructed large-scale database which consists of more than 3.3 million names. The database contains more than 105 thousand unique names with a total of more than 1.1 million images of personal names, which proves useful for transfer learning to other settings. We provide three examples hereof, obtaining significantly improved transcription accuracy on both Danish and US census data. In addition, we present benchmark results for deep learning models automatically transcribing the personal names from the scanned documents. Through making more challenging large-scale databases publicly available we hope to foster more sophisticated, accurate, and robust models for handwritten text recognition. [ABSTRACT FROM AUTHOR]

In the field of regulatory science, reviewing literature is an essential and important step, which most of the time is conducted by manually reading hundreds of articles. Although this process is highly time-consuming and labor-intensive, most output of this process is not well transformed into machine-readable format. The limited availability of data has largely constrained the artificial intelligence (AI) system development to facilitate this literature reviewing in the regulatory process. In the past decade, AI has revolutionized the area of text mining as many deep learning approaches have been developed to search, annotate, and classify relevant documents. After the great advancement of AI algorithms, a lack of high-quality data instead of the algorithms has recently become the bottleneck of AI system development. Herein, we constructed two large benchmark datasets, Chlorine Efficacy dataset (CHE) and Chlorine Safety dataset (CHS), under a regulatory scenario that sought to assess the antiseptic efficacy and toxicity of chlorine. For each dataset, ∼10,000 scientific articles were initially collected, manually reviewed, and their relevance to the review task were labeled. To ensure high data quality, each paper was labeled by a consensus among multiple experienced reviewers. The overall relevance rate was 27.21% (2,663 of 9,788) for CHE and 7.50% (761 of 10,153) for CHS, respectively. Furthermore, the relevant articles were categorized into five subgroups based on the focus of their content. Next, we developed an attention-based classification language model using these two datasets. The proposed classification model yielded 0.857 and 0.908 of Area Under the Curve (AUC) for CHE and CHS dataset, respectively. This performance was significantly better than permutation test (p < 10E-9), demonstrating that the labeling processes were valid. To conclude, our datasets can be used as benchmark to develop AI systems, which can further facilitate the literature review process in regulatory science. • Two Benchmark labeled datasets for literature review were developed. • The labeling progress involved multiple reviewers to arrive at consensus of the final result. • A down-stream analysis with language modeling was applied to demonstrate the validity of the benchmark datasets. [ABSTRACT FROM AUTHOR]

Deep neural network models with strong feature extraction capacity are prone to overfitting and fail to adapt quickly to new tasks with few samples. Gradient-based meta-learning approaches can minimize overfitting and adapt to new tasks fast, but they frequently use shallow neural networks with limited feature extraction capacity. We present a simple and effective approach called Meta-Transfer-Adjustment learning (MTA) in this paper, which enables deep neural networks with powerful feature extraction capabilities to be applied to few-shot scenarios while avoiding overfitting and gaining the capacity for quickly adapting to new tasks via training on numerous tasks. Our presented approach is classified into two major parts, the Feature Adjustment (FA) module, and the Task Adjustment (TA) module. The feature adjustment module (FA) helps the model to make better use of the deep network to improve feature extraction, while the task adjustment module (TA) is utilized for further improve the model's fast response and generalization capabilities. The proposed model delivers good classification results on the benchmark small sample datasets MiniImageNet and Fewshot-CIFAR100, as proved experimentally. [ABSTRACT FROM AUTHOR]

• An efficient non-uniform motion deblurring network is proposed to introduce edge information to guide image deblurring task. • A new Multi-stream Adaptive Fusion Module (MAFM) is designed to focus on crucial information in different information flows. • A novel Dense Attention Feature Extraction Module (DAFEM) is developed to obtain significant recovery information. • Edge loss and perceptual loss are introduced to optimize the generator to improve the image restoration effect. Non-uniform motion deblurring has been a challenging problem in the field of computer vision. Currently, deep learning-based deblurring methods have made promising achievements. In this paper, we propose a new joint strong edge and multi-stream adaptive fusion network to achieve non-uniform motion deblurring. The edge map and the blurred map are jointly used as network inputs and Edge Extraction Network (EEN) guides the Deblurring Network (DN) for image recovery and to complement the important edge information. The Multi-stream Adaptive Fusion Module (MAFM) adaptively fuses the edge information and features from the encoder and decoder to reduce feature redundancy to avoid image artifacts. Furthermore, the Dense Attention Feature Extraction Module (DAFEM) is designed to focus on the severely blurred regions of blurry images to obtain important recovery information. In addition, an edge loss function is added to measure the difference of edge features between the generated and clear images to further recover the edges of the deblurred images. Experiments show that our method outperforms currently public methods in terms of PSNR, SSIM and VIF, and generates images with less blur and sharper edges. [ABSTRACT FROM AUTHOR]

In this paper we propose an artificial intelligence imaging solution to support archaeologists in the classification task of Roman commonware potsherds. Usually, each potsherd is represented by its sectional profile as a two dimensional black–white image and printed in archaeological books related to specific archaeological excavations. The partiality and handcrafted variance of the fragments make their matching a challenging problem: we propose to pair similar profiles via the unsupervised hierarchical clustering of non-linear features learned in the latent space of a deep convolutional Variational Autoencoder (VAE) network. Our contribution also include the creation of a ROman COmmonware POTtery (ROCOPOT) database, with more than 4000 potsherds profiles extracted from 25 Roman pottery corpora, and a MATLAB GUI software for the easy inspection of shape similarities. Results are commented both from a mathematical and archaeological perspective so as to unlock new research directions in both communities. • A database of Roman commonware pottery profiles from Tyrrhenian Italy is produced. • A deep learning algorithm supports the clustering of Roman commonware profiles. • The review of profile matching is eased by a dedicated visual software. • The probability of detecting new or unseen profile matches is increased. • Unveiling new profile matches broaden the knowledge on excavation sites. [ABSTRACT FROM AUTHOR]

Recognition of construction waste compositions using computer vision (CV) is increasingly explored to enable its subsequent management, e.g., determining chargeable levy at disposal facilities or waste sorting using robot arms. However, the applicability of existing CV-enabled construction waste recognition in real-life scenarios is limited by their relatively low accuracy, characterized by a failure to distinguish boundaries among different waste materials. This paper aims to propose a novel boundary-aware Transformer (BAT) model for fine-grained composition recognition of construction waste mixtures. First, a pre-processing workflow is devised to separate the hard-to-recognize edges from the background. Second, a Transformer structure with a self-designed cascade decoder is developed to segment different waste materials from construction waste mixtures. Finally, a learning-enabled edge refinement scheme is used to fine-tune the ignored boundaries, further boosting the segmentation precision. The performance of the BAT model was evaluated on a benchmark dataset comprising nine types of materials in a cluttered and mixture state. It recorded a 5.48% improvement of MIoU (mean intersection over union) and 3.65% of MAcc (Mean Accuracy) against the baseline. The research contributes to the body of interdisciplinary knowledge by presenting a novel deep learning model for construction waste material semantic segmentation. It can also expedite the applications of CV in construction waste management to achieve a circular economy. • A boundary-aware transformer (BAT) model is proposed for CW composition recognition. • The BAT model innovatively integrates Transformer with a cascade decoder. • An edge refinement network is trained to finetune the ignored boundaries. • BAT outperforms the baseline by 5.48% for mean intersection over union (MIoU). [ABSTRACT FROM AUTHOR]