A Multi-Branch Dual Residual Attention CNN Model for Pediatric Pneumonia Classification

Describing what has changed in a scene can be useful to a user, but only if generated text focuses on what is semantically relevant. It is thus important to distinguish distractors (e.g. a viewpoint change) from relevant changes (e.g. an object has moved). We present a novel Dual Dynamic Attention Model (DUDA) to perform robust Change Captioning. Our model learns to distinguish distractors from semantic changes, localize the changes via Dual Attention over “before” and “after” images, and accurately describe them in natural language via Dynamic Speaker, by adaptively focusing on the necessary visual inputs (e.g. “before” or “after” image). To study the problem in depth, we collect a CLEVR-Change dataset, built off the CLEVR engine, with 5 types of scene changes. We benchmark a number of baselines on our dataset, and systematically study different change types and robustness to distractors. We show the superiority of our DUDA model in terms of both change captioning and localization. We also show that our approach is general, obtaining state-of-the-art results on the recent realistic Spot-the-Diff dataset which has no distractors.

The heart is a very crucial organ of the body. Concerted efforts are constantly put forward to provide adequate monitoring of the heart. A heart disorder is reported to cause a lot of hidden ailments resulting in numerous deaths. Early heart monitoring using an electrocardiogram (ECG) through the advancement of computer-aided diagnostic (CAD) systems is widely used. Meanwhile, the use of human reading of ECG results are faced with many challenges of inaccurate and unreliable interpretations. Over two decades, studies provided artificial intelligence (AI) technique using machine learning (ML) algorithms as a fast and reliable technique for ECG heartbeat classification. Moreover, in recent times, deep learning (DL) techniques have been focused on providing automatic feature extraction and better classification performance. On the other hand, the challenge with the ECG data is its imbalance nature. Therefore, this paper proposes a cost-based dual convolutional attention transfer DL model for ECG classification. The proposed model uses PhysionNet-MIT-BIH and Physikalisch-Technische Bundesanstalt (PTB) Diagnostics datasets. The first part uses the MIT-BIH for ECG categorization, while representations learned from the first classifier are used for PTB analysis through transfer learning (TL). The proposed model is evaluated and compared with well-performing conventional ML models based on their F1-score and accuracy scores. Our experimental finding show that the proposed model outperformed the well-performing ML models as well as competitive with past studies for both the classification and TL part, having obtained 98.45% for both F1-score and accuracy. The proposed model is applicable to real-life trials and experiments for ECG heartbeat and other similar domains.

Decoding the scientific creative-ability of subjects using dual attention induced graph convolutional-capsule network

Accurate wound segmentation is crucial for the precise diagnosis and treatment of various skin conditions through image analysis. In this paper, we introduce a novel dual attention U-Net model designed for precise wound segmentation. Our proposed architecture integrates two widely used deep learning models, VGG16 and U-Net, incorporating dual attention mechanisms to focus on relevant regions within the wound area. Initially trained on diabetic foot ulcer images, we fine-tuned the model to acute and chronic wound images and conducted a comprehensive comparison with other state-of-the-art models. The results highlight the superior performance of our proposed dual attention model, achieving a Dice coefficient and IoU of 94.1% and 89.3%, respectively, on the test set. This underscores the robustness of our method and its capacity to generalize effectively to new data.

Aim/Purpose: To optimize healthcare resource allocation using residual convolutional neural networks. Background: In the early stages, several traditional methods were adopted and implement-ed; however, the rise of AI and its technologies increased development in the healthcare sector and made it reach a better height in Industry 4.0. In the early stages, several traditional methods were adopted and implemented; however, the rise of AI and its technologies increased development in the healthcare sector and made it reach a better height in Industry 4.0. The main problem of this research is focusing on the inefficient allocation of healthcare resources, which leads to less outcome and accuracy. This research’s main novelty and objective is to implement a predictive model that may allocate resources based on several factors. Methodology: In the proposed method, Residual CNNs, a deep learning architecture well-known for its efficacy in image classification applications, we assess healthcare data and estimate ideal resource distribution. Residual CNNs are well-trained in the dataset on several factors and characteristics. The model produces predictions of resource allocation that maximize healthcare outcomes using comprehension of complex relationships and patterns in the data. Contribution: The novel feature of this work is the integration of the state-of-the-art deep learning architecture Residual CNNs into the domain of healthcare resource allocation. The proposed method, Residual CNNs, is well-trained in the dataset on several factors and characteristics. The model produces predictions of resource allocation that maximize healthcare outcomes by comprehending complex relationships and patterns in the data. Findings: We show experimentally that the proposed approach effectively allocates healthcare resources. The residual CNN model outperforms traditional methods in accurately predicting resource allocation needs across different regions and demographic groups. We find significant increases in resource allocation efficiency by applying deep learning techniques, which enhance healthcare outcomes and reduce treatment disparities. Recommendation for Researchers: Investigations should prioritize the validation of the algorithm in various healthcare environments to assess its efficacy in clinical application. Future Research: This work can be enhanced in future research using several deep-learning algorithms to achieve better accuracy and performance.

Oil spills cause tremendous damage to marine, coastal environments, and ecosystems. Previous deep learning-based studies have addressed the task of detecting oil spills as a semantic segmentation problem. However, further improvement is still required to address the noisy nature of the Synthetic Aperture Radar (SAR) imagery problem, which limits segmentation performance. In this study, a new deep learning model based on the Dual Attention Model (DAM) is developed to automatically detect oil spills in a water body. We enhanced a conventional UNet segmentation network by integrating a dual attention model DAM to selectively highlight the relevant and discriminative global and local characteristics of oil spills in SAR imagery. DAM is composed of a Channel Attention Map and a Position Attention Map which are stacked in the decoder network of UNet. The proposed DAM-UNet is compared with four baselines, namely fully convolutional network, PSPNet, LinkNet, and traditional UNet. The proposed DAM-UNet outperforms the four baselines, as demonstrated empirically. Moreover, the EG-Oil Spill dataset includes a large set of SAR images with 3000 image pairs. The obtained overall accuracy of the proposed method increased by 3.2% and reaches 94.2% compared with that of the traditional UNet. The study opens new development ideas for integrating attention modules into other deep learning tasks, including machine translation, image-based analysis, action recognition, and speech recognition.

Dual attention and dual fusion: An accurate way of image-based geo-localization

In the application of remote sensing image (RSI) scene classification, in order to solve the contradiction between the accuracy of Convolutional Neural Network (CNN) and the large amount of model parameters, a novel dual knowledge distillation (DKD) model combining dual attention (DA) and spatial structure (SS) is designed. First, new DA and SS modules are constructed and introduced into ResNet101 and light-weight CNN designed as teacher and student networks respectively. Then, in order to improve its local feature extraction and high-level semantic representation abilities for RSI by transmission the DA and SS knowledge in the teacher network to the student network, we design the corresponding DA and SS distillation losses. The comparative experimental results based on AID and NWPU-45 datasets show that when the training ratio is 20%, the accuracy of the student network after DKD is improved by 7.57% and 7.28% respectively, and in the case of fewer parameters, DKD has higher accuracy than most other methods.

Nowadays, session-based recommendation plays an increasingly important role in the e-commerce field, which predicts the item that a user may click next time based on the sequence of user clicks. However, in real-world scenarios, due to various factors, there is noise in the process of user click behavior. For example, an unexpected click may not be the user's true intention and thus affect the user's behavior prediction. The current denoising methods have the following challenges: denoising directly from a single user's click sequence is not sufficient, the impact of unexpected clicks is not fully considered, and the additional information of other users is not fully utilized. To address these challenges, a new denoising dual sparse graph attention model for session-based recommendation abbreviated as DDSG, which not only considers the information in the current session, but also utilizes the information outside the current session. In the current session, this paper uses position encoding and gated neural networks that are biased towards frequency information to obtain the initial embedding, and uses the self-attention mechanism to model the target representation. In terms of denoising, we first iteratively denoise the representation obtained in the session, and perform sparse self-attention denoising on the session representation based on the target representation. Outside the current session, we take other sessions with similar interests to the current session to enhance the current session. Finally, the user's next click item is predicted by combining internal and extra information of the session. Experiments on three e-commerce datasets demonstrate our model exceeded the optimal SOTA model by 107%, achieving the highest performance and verifying the effectiveness of our model.

Recognition of Handwritten Mathematical Expressions (HMEs) is a challenging problem because of the complicated structure and uncommon math symbols contained in HMEs. Moreover, the lack of training data is a serious issue, especially for deep learning-based systems. In this paper, we proposed a dual loss attention model that utilizes the existing latex corpus to improve accuracy. The proposed dual loss attention has two losses, including decoder loss and context matching loss to learn semantic invariant features for the encoder and latex grammar for the decoder from handwritten and printed MEs. The results of experiments on the CROHME 2014 and 2016 databases demonstrate the superiority and effectiveness of our proposed model. These results are competitive compared to others reported in recent literature.

In recent years, both the academic and commercial communities have paid great attentions on embedding methods to analyze all kinds of network data. Despite of the great successes of DeepWalk and the following neural models, only a few of them have the ability to incorporate contents and labels into low-dimensional representation vectors of nodes. Besides, most network embedding methods only consider universal representations and the optimal representations could not be learned for specific tasks. In this paper, we propose a Multi-task Dual Attention LSTM model (dubbed as MDAL), which can capture structure, content, and label information of network and adjust representation vectors according to the concrete downstream task simultaneously. For the target node, MDAL leverages Tree-LSTM structure to extract structure, text and label information from its neighborhood. With the help of dual attention mechanism, the content related and label related neighbor nodes are emphasized during embedding. MDAL utilizes a multi-task learning framework that considering both network embedding and downstream tasks. The appropriate loss functions are proposed for task adaption and a joint optimization process is conducted for task-specific network embedding. We compare MDAL with the state-of-the-art and strong baselines for node classification, network visualization and link prediction tasks. Experimental results show the effectiveness and superiority of our proposed MDAL model.

In this article, classification problems for a class of tensor datasets with a noisy environment are investigated. To address such issues, a novel fuzzy support tensor machine (FSTM) dual model with robustness is established. First, for each input sample in the noisy tensor dataset, we define three kinds of fuzzy membership functions, such as linear, cosine, and exponential forms. In particular, the reconstruction process from one-dimensional (1-D) vector data to third-order tensor data is also derived in the Appendix. Second, the original optimization model of an FSTM on fuzzy membership is designed by constructing the vector pattern of the traditional support vector machine (SVM) models into a tensor pattern. Next, by introducing the Lagrangian multiplier method and tensor-Tucker decomposition method to the original FSTM model, an FSTM dual model without tensor inner product operation is obtained for the first time. Such a dual model with tensor-Tucker decomposition form can avoid conservativeness caused by the vectorization of tensor data in the traditional SVM model. Furthermore, an FSTM classifier is derived by the designed numerical algorithm, and the classification generalization error bound of the FSTM model with a general form is developed. It is worth noting that a linear least-squares FSTM (LLS-FSTM) equation with tensor-Tucker decomposition is also designed in the Appendix to further reduce the slightly time-consuming problem of the solving the dual optimization model FSTM. Finally, two numerical examples are presented to verify the feasibility and validity of the derived FSTM classifier.

Explainable detection of myocardial infarction using deep learning models with Grad-CAM technique on ECG signals

Dual attention model with reinforcement learning for classification of histology whole-slide images

Audio-Visual Speech and Gesture Recognition using Dual Sampling Residual Attention CNN on Mobile Devices