Attention Map Research Articles

Predicting enzymatic function of protein sequences with attention.

There is a growing number of available protein sequences, but only a limited amount has been manually annotated. For example, only 0.25% of all entries of UniProtKB are reviewed by human annotators. Further developing automatic tools to infer protein function from sequence alone can alleviate part of this gap. In this article, we investigate the potential of Transformer deep neural networks on a specific case of functional sequence annotation: the prediction of enzymatic classes. We show that our EnzBert transformer models, trained to predict Enzyme Commission (EC) numbers by specialization of a protein language model, outperforms state-of-the-art tools for monofunctional enzyme class prediction based on sequences only. Accuracy is improved from 84% to 95% on the prediction of EC numbers at level two on the EC40 benchmark. To evaluate the prediction quality at level four, the most detailed level of EC numbers, we built two new time-based benchmarks for comparison with state-of-the-art methods ECPred and DeepEC: the macro-F1 score is respectively improved from 41% to 54% and from 20% to 26%. Finally, we also show that using a simple combination of attention maps is on par with, or better than, other classical interpretability methods on the EC prediction task. More specifically, important residues identified by attention maps tend to correspond to known catalytic sites. Quantitatively, we report a max F-Gain score of 96.05%, while classical interpretability methods reach 91.44% at best. Source code and datasets are respectively available at https://gitlab.inria.fr/nbuton/tfpc and https://doi.org/10.5281/zenodo.7253910.

S 2 CL-Leaf Net : Recognizing Leaf Images Like Human Botanists

Automatically classifying plant leaves is a challenging fine-grained classification task because of the diversity in leaf morphology, including size, texture, shape, and venation. Although powerful deep learning-based methods have achieved great improvement in leaf classification, these methods still require a large number of well-labeled samples for supervised training, which is difficult to get. In contrast, relying on the specific coarse-to-fine classification strategy, human botanists only require a small number of samples for accurate leaf recognition. Inspired by the classification strategy of human botanists, we propose a novel S 2 CL-Leaf Net , which exploits multi-granularity clues with a hierarchical attention mechanism and boosts the learning ability with the supervised sampling contrastive learning with limited training samples to classify plant leaves as human botanists do. Specifically, to fully explore and exploit the subtle details of the leaves, a novel sampling transformation mechanism is combined with the supervised contrastive learning to enhance the network’s perception of details by amplifying the discriminative regions with a weighted sampling of different regions. Furthermore, we construct the hierarchical attention mechanism to produce attention maps of different granularity, which helps to discover details in leaves that are important for classification. Experiments are conducted on the open-access leaf datasets, including Flavia, Swedish, and LeafSnap, which prove the effectiveness of the proposed S 2 CL-Leaf Net .

Attention-based dual-color space fusion network for low-light image enhancement

The low visibility and dull colors associated with low-light images are not only difficult to satisfy the photographer, but also hinder further visual tasks. In this study, we propose an attention-based dual-color space fusion network to enhance low-light images. By introducing HSV color space, the network can solve the problems of dim color and insufficient contrast existing in previous RGB color space methods. In both color spaces, the network adopts a large number of attention mechanisms to highlight important features in real time. To better achieve feature fusion across color space, we design adaptive large-kernel attention (ALKA) and feature extraction module (FEM), respectively. The ALKA adaptively selects features for enriching the encoding input in the HSV color space during the RGB color space encoding process. The FEM assists the network in more comprehensive supervision of the enhancement process and generates the attention map to transmit the marking features in the previous color space. At the end of the network, we also design channel enhancement module (CEM) to protect the texture details lost during the scale change. We have conducted experiments on a variety of public datasets, and both subjective and objective comparisons have demonstrated the particularity and prominence of our method.

Pairwise Guided Multilayer Cross-Fusion Network for Bird Image Recognition

Bird identification is the first step in collecting data on bird diversity and abundance, which also helps research on bird distribution and population measurements. Most research has built end-to-end training models for bird detection task via CNNs or attentive models, but many perform unsatisfactorily in fine-grained bird recognition. Bird recognition tasks are highly influenced by factors, including the similar appearance of different subcategories, diverse bird postures, and other interference factors such as tree branches and leaves from the background. To tackle this challenge, we propose the Progressive Cross-Union Network (PC-Net) to capture more subtle parts with low-level attention maps. Based on cross-layer information exchange and pairwise learning, the proposed method uses two modules to improve feature representation and localization. First, it utilizes low- and high-level information for cross-layer feature fusion, which enables the network to extract more comprehensive and discriminative features. Second, the network incorporates deep semantic localization to identify and enhance the most relevant regions in the images. In addition, the network is designed with a semantic guidance loss to improve its generalization for variable bird poses. The PC-Net was evaluated on an extensively used birds dataset (CUB-200-2011), which contains 200 birds subcategories. The results demonstrate that the PC-Net achieved an impressive recognition accuracy of 89.2%, thereby outperforming maintained methods in bird subcategory identification. We also achieved competitive results on two other datasets with data on cars and airplanes. The results indicated that the PC-Net improves the accuracy of diverse bird recognition, as well as other fine-grained recognition scenarios.

How to identify pollen like a palynologist: A prior knowledge-guided deep feature learning for real-world pollen classification

Airborne pollen identification is crucial to help patients prevent pollinosis symptoms. Existing data-driven methods rely on large-scale pollen images with simple backgrounds. In real scenarios, the background is complex and the data scale is small. Therefore, these methods suffer from two challenges: (1) Irrelevant information interference; (2) Incomplete feature attention. To overcome these challenges, we propose a prior knowledge-guided deep feature learning (PK-DFL) for real-world optical microscope image classification. Its main steps are as follows: Pollen location is designed to locate pollen grains based on color features, aiming to boost the accuracy of shape and texture prior feature extraction. Shape-texture awareness helps to extract the shape and texture of pollen grains via predefined feature extractors (i.e., a set of shape descriptors and an improved SFTA). These features are used to construct two types of prior knowledge, namely shape-texture attention maps (STA maps) and shape-texture feature vectors (STF vectors). Pollen classification uses a deep network (CNN) to classify pollen via imitating the pollen identification procedure of palynologists. It uses STA maps to weight pollen images and convolutional feature maps for instructing the CNN to focus on critical areas of pollen images (for the first challenge). STF vectors are employed to obtain the inter-class similarity of pollen via template matching. This information is further converted to soft targets that are used to supervise the CNN attending to comprehensive key features (for the second challenge). Extensive experiments on real-world datasets demonstrate the effectiveness of our PK-DFL (with accuracy and F1-score over 88%).

Contour attention network for cerebrovascular segmentation from TOF-MRA volumetric images.

Cerebrovascular segmentation is a crucial step in the computer-assisted diagnosis of cerebrovascular pathologies. However, accurate extraction of cerebral vessels from time-of-flight magnetic resonance angiography (TOF-MRA) data is still challenging due to the complex topology and slender shape. The existing deep learning-based approaches pay more attention to the skeleton and ignore the contour, which limits the segmentation performance of the cerebrovascular structure. We aim to weight the contour of brain vessels in shallow features when concatenating with deep features. It helps to obtain more accurate cerebrovascular details and narrows the semantic gap between multilevelfeatures. This work proposes a novel framework for priority extraction of contours in cerebrovascular structures. We first design a neighborhood-based algorithm to generate the ground truth of the cerebrovascular contour from original annotations, which can introduce useful shape information for the segmentation network. Moreover, We propose an encoder-dual decoder-based contour attention network (CA-Net), which consists of the dilated asymmetry convolution block (DACB) and the Contour Attention Module (CAM). The ancillary decoder uses the DACB to obtain cerebrovascular contour features under the supervision of contour annotations. The CAM transforms these features into a spatial attention map to increase the weight of the contour voxels in main decoder to better restored the vessel contourdetails. The CA-Net is thoroughly validated using two publicly available datasets, and the experimental results demonstrate that our network outperforms the competitors for cerebrovascular segmentation. We achieved the average dice similarity coefficient ( ) of 68.15 and 99.92% in natural and synthetic datasets. Our method segments cerebrovascular structures with bettercompleteness. We propose a new framework containing contour annotation generation and cerebrovascular segmentation network that better captures the tiny vessels and improve vessel connectivity.

Adopting multiple vision transformer layers for fine-grained image representation

Accurate discriminative regions proposal has an important effect for fine-grained image recognition. The vision transformer (ViT) brings about a striking effect in computer vision duo to its innate muti-head self-attention mechanism. However, the attention maps are gradually similar after certain layers and since ViT adds classification token for perform classification, it is unable to effectively select discriminative image patches for fine-grained image classification. To accurately detect discriminative regions, we propose a novel network AMTrans, which efficiently increases layers to learn diverse features and utilizes integrated raw attention maps to capture more salient feature. Specifically, we employ DeepViT as backbone to solve the attention collapse issue. Then, we fuse each head attention weight within each layer to produce attention weight map. After that, we alternatively use recurrent residual refinement blocks to promote salient feature detection and then utilize semantic grouping method to propose the discriminative feature region. A lot of experiments prove that AMTrans acquires the SOTA performance on three widely used fine-grained datasets under the same settings, involving Stanford-Cars, Stanford-Dogs and CUB-200-2011.

One Model to Synthesize Them All: Multi-Contrast Multi-Scale Transformer for Missing Data Imputation.

Multi-contrast magnetic resonance imaging (MRI) is widely used in clinical practice as each contrast provides complementary information. However, the availability of each imaging contrast may vary amongst patients, which poses challenges to radiologists and automated image analysis algorithms. A general approach for tackling this problem is missing data imputation, which aims to synthesize the missing contrasts from existing ones. While several convolutional neural networks (CNN) based algorithms have been proposed, they suffer from the fundamental limitations of CNN models, such as the requirement for fixed numbers of input and output channels, the inability to capture long-range dependencies, and the lack of interpretability. In this work, we formulate missing data imputation as a sequence-to-sequence learning problem and propose a multi-contrast multi-scale Transformer (MMT), which can take any subset of input contrasts and synthesize those that are missing. MMT consists of a multi-scale Transformer encoder that builds hierarchical representations of inputs combined with a multi-scale Transformer decoder that generates the outputs in a coarse-to-fine fashion. The proposed multi-contrast Swin Transformer blocks can efficiently capture intra- and inter-contrast dependencies for accurate image synthesis. Moreover, MMT is inherently interpretable as it allows us to understand the importance of each input contrast in different regions by analyzing the in-built attention maps of Transformer blocks in the decoder. Extensive experiments on two large-scale multi-contrast MRI datasets demonstrate that MMT outperforms the state-of-the-art methods quantitatively and qualitatively.

DAUF: A disease-related attentional UNet framework for progressive and stable mild cognitive impairment identification

Identifying progressive mild cognitive impairment (pMCI) and stable mild cognitive impairment (sMCI) plays a significant role in early Alzheimer's disease (AD) diagnosis, which can effectively boost the life quality of patients. Recently, convolutional neural network (CNN)- based methods using structural magnetic resonance imaging (sMRI) images have shown effective for AD identification. However, these CNN-based methods fail to effectively explore the feature extraction of disease-related multi-scale tissues, such as ventricles, hippocampi and cerebral cortex. To address this issue, we propose an end-to-end disease-related attentional UNet framework (DAUF) for identifying pMCI and sMCI, by embedding a devised dual disease-related attention module (D2AM) and a novel tree-structured feature fusion classifier (TFFC). Specifically, D2AM leverages the complementarity between feature maps and attention maps and the complementary features from the encoder and decoder, so as to highlight discriminative semantic and detailed features. Additionally, TFFC is a powerfully joint multi-scale feature fusion and classification head, by employing the homogeneity among multi-scale features, so that the discriminative features of the multi-scale tissues are adequately fused for enhancing classification performance. Finally, extensive experiments demonstrate the superior performance of DAUF, with the effectiveness of D2AM and TFFC on identifying pMCI and sMCI subjects.

Artificial Intelligence to Preoperatively Predict Proximal Junction Kyphosis Following Adult Spinal Deformity Surgery: Soft Tissue Imaging May Be Necessary for Accurate Models.

Retrospective cohort. In a cohort of patients undergoing adult spinal deformity (ASD) surgery, we used artificial intelligence to compare three models of preoperatively predicting radiographic proximal junction kyphosis (PJK) using: (1) traditional demographics and radiographic measurements, (2) raw preoperative scoliosis radiographs, and (3) raw preoperative thoracic magnetic resonance imaging (MRI). Despite many proposed risk factors, PJK following ASD surgery remains difficult to predict. A single-institution, retrospective cohort study was undertaken for patients undergoing ASD surgery from 2009 to 2021. PJK was defined as a sagittal Cobb angle of upper-instrumented vertebra (UIV) and UIV+2>10° and a postoperative change in UIV/UIV+2>10°. For model 1, a support vector machine was used to predict PJK within 2 years postoperatively using clinical and traditional sagittal/coronal radiographic variables and intended levels of instrumentation. Next, for model 2, a convolutional neural network (CNN) was trained on raw preoperative lateral and posterior-anterior scoliosis radiographs. Finally, for model 3, a CNN was trained on raw preoperative thoracic T1 MRIs. A total of 191 patients underwent ASD surgery with at least 2-year follow-up and 89 (46.6%) developed radiographic PJK within 2 years. Model 1: Using clinical variables and traditional radiographic measurements, the model achieved a sensitivity: 57.2% and a specificity: 56.3%. Model 2: a CNN with raw scoliosis x-rays predicted PJK with a sensitivity: 68.2% and specificity: 58.3%. Model 3: a CNN with raw thoracic MRIs predicted PJK with average sensitivity: 73.1% and specificity: 79.5%. Finally, an attention map outlined the imaging features used by model 3 elucidated that soft tissue features predominated all true positive PJK predictions. The use of raw MRIs in an artificial intelligence model improved the accuracy of PJK prediction compared with raw scoliosis radiographs and traditional clinical/radiographic measurements. The improved predictive accuracy using MRI may indicate that PJK is best predicted by soft tissue degeneration and muscle atrophy.

Aquaculture defects recognition via multi-scale semantic segmentation

Aquaculture net pen defects such as biofouling, vegetation, and holes are key challenges to efficient and sustainable fish production in aquaculture. These defects must be monitored to avoid problems with aquaculture net safety as well as fish growth. Recently, deep learning methods have been adopted to solve detection and classification problems in different applications. However, the conventional methods are challenging to meet the demands of high precision and real-time detection of aquaculture net defects in a complex marine environment. Towards this end, this paper proposes an autonomous net pen defect detection system that contains a novel multi-scale semantic segmentation topology for detecting the biofouling, vegetation, and hole problems in the aquaculture environment. In particular, we emphasize fusing the attention maps obtained across different decomposition levels of the network to generate rich feature distributions that enable the accurate extraction of biofouling, vegetation, and holes within aquaculture nets. Moreover, the proposed model is thoroughly tested on two private datasets and two public datasets which were acquired from a real-field fish farms. Across all four datasets, the proposed framework showed remarkable biofouling, vegetation, and hole detection performance, where it outperformed state-of-the-art methods by 6.58%, 3.69%, 6.44%, and 4.78% in terms of mean average precision across LABUST, KU, NDv1, and NDv2 datasets, respectively.

Word self-update contrastive adversarial networks for text-to-image synthesis

Synthesizing realistic fine-grained images from text descriptions is a significant computer vision task. Although many GANs-based methods have been proposed to solve this task, generating high-quality images consistent with text information remains a difficult problem. These existing GANs-based methods ignore important words due to the use of fixed initial word features in generator, and neglect to learn semantic consistency between images and texts for discriminators. In this article, we propose a novel attentional generation and contrastive adversarial framework for fine-grained text-to-image synthesis, termed as Word Self-Update Contrastive Adversarial Networks (WSC-GAN). Specifically, we introduce a dual attention module for modeling color details and semantic information. With a new designed word self-update module, the generator can leverage visually important words to compute attention maps in the feature synthesis module. Furthermore, we contrive multi-branch contrastive discriminators to maintain better consistency between the generated image and text description. Two novel contrastive losses are proposed for our discriminators to impose image-sentence and image-word consistency constraints. Extensive experiments on CUB and MS-COCO datasets demonstrate that our method achieves better performance compared with state-of-the-art methods.

Sparse self-attention transformer for image inpainting

Learning-based image inpainting methods have made remarkable progress in recent years. Nevertheless, these methods still suffer from issues such as blurring, artifacts, and inconsistent contents. The use of vanilla convolution kernels, which have limited perceptual fields and spatially invariant kernel coefficients, is one of the main causes for these problems. In contrast, the multi-headed attention in the transformer can effectively model non-local relations among input features by generating adaptive attention scores. Therefore, this paper explores the feasibility of employing the transformer model for the image inpainting task. However, the multi-headed attention transformer blocks pose a significant challenge due to their overwhelming computational cost. To address this issue, we propose a novel U-Net style transformer-based network for the inpainting task, called the sparse self-attention transformer (Spa-former). The Spa-former retains the long-range modeling capacity of transformer blocks while reducing the computational burden. It incorporates a new channel attention approximation algorithm that reduces attention calculation to linear complexity. Additionally, it replaces the canonical softmax function with the ReLU function to generate a sparse attention map that effectively excludes irrelevant features. As a result, the Spa-former achieves effective long-range feature modeling with fewer parameters and lower computational resources. Our empirical results on challenging benchmarks demonstrate the superior performance of our proposed Spa-former over state-of-the-art approaches.

ESA: Excitation-Switchable Attention for convolutional neural networks

Although various attention mechanisms can boost the representational power of convolutional neural networks (CNNs) and improve their performance, selecting an appropriate attention module becomes challenging when backbones or datasets change. Besides, as different CNN layers can learn distinct semantic features, applying the same attention module across all layers may not yield optimal results for enhancing the performance of a deep neural network. To address the above issues, we propose a novel excitation-switchable attention (ESA) to automatically select and integrate different excitation modules to compute attention maps, enabling a DNN to apply different attention modules in different layers for better feature learning and performance improvement. Extensive experiments on three widely-used image classification benchmarks demonstrate the superiority of our ESA over several well-known and widely-adopted attention modules.

Comprehensive fully-automatic multi-depth grading of the clinical types of macular neovascularization in OCTA images

AbstractOptical Coherence Tomography Angiography or OCTA represents one of the main means of diagnosis of Age-related Macular Degeneration (AMD), the leading cause of blindness in developed countries. This eye disease is characterized by Macular Neovascularization (MNV), the formation of vessels that tear through the retinal tissues. Four types of MNV can be distinguished, each representing different levels of severity. Both the aggressiveness of the treatment and the recovery of the patient rely on an early detection and correct diagnosis of the stage of the disease. In this work, we propose the first fully-automatic grading methodology that considers all the four clinical types of MNV at the three most relevant OCTA scanning depths for the diagnosis of AMD. We perform both a comprehensive ablation study on the contribution of said depths and an analysis of the attention maps of the network in collaboration with experts of the domain. Our proposal aims to ease the diagnosis burden and decrease the influence of subjectivity on it, offering a explainable grading through the visualization of the attention of the expert models. Our grading proposal achieved satisfactory results with an AUC of 0.9224 ± 0.0381. Additionally, the qualitative analysis performed in collaboration with experts revealed the relevance of the avascular plexus in the grading of all three types of MNV (despite not being directly involved in some of them). Thus, our proposal is not only able to robustly detect MNV in complex scenarios, but also aided to discover previously unconsidered relationships between plexuses.

Deep neural network-based clustering of deformation curves reveals novel disease features in PLN pathogenic variant carriers

Echocardiographic deformation curves provide detailed information on myocardial function. Deep neural networks (DNNs) may enable automated detection of disease features in deformation curves, and improve the clinical assessment of these curves. We aimed to investigate whether an explainable DNN-based pipeline can be used to detect and visualize disease features in echocardiographic deformation curves of phospholamban (PLN) p.Arg14del variant carriers. A DNN was trained to discriminate PLN variant carriers (n = 278) from control subjects (n = 621) using raw deformation curves obtained by 2D-speckle tracking in the longitudinal axis. A visualization technique was used to identify the parts of these curves that were used by the DNN for classification. The PLN variant carriers were clustered according to the output of the visualization technique. The DNN showed excellent discriminatory performance (C-statistic 0.93 [95% CI 0.87–0.97]). We identified four clusters with PLN-associated disease features in the deformation curves. Two clusters showed previously described features: apical post-systolic shortening and reduced systolic strain. The two other clusters revealed novel features, both reflecting delayed relaxation. Additionally, a fifth cluster was identified containing variant carriers without disease features in the deformation curves, who were classified as controls by the DNN. This latter cluster had a very benign disease course regarding development of ventricular arrhythmias. Applying an explainable DNN-based pipeline to myocardial deformation curves enables automated detection and visualization of disease features. In PLN variant carriers, we discovered novel disease features which may improve individual risk stratification. Applying this approach to other diseases will further expand our knowledge on disease-specific deformation patterns.Graphical abstractOverview of the deep neural network-based pipeline for feature detection in myocardial deformation curves. Firstly, phospholamban (PLN) p.Arg14del variant carriers and controls were selected and a deep neural network (DNN) was trained to detect the PLN variant carriers. Subsequently, a clustering-based approach was performed on the attention maps of the DNN, which revealed 4 distinct phenotypes of PLN variant carriers with different prognoses. Moreover, a cluster without features and a benign prognosis was detected.

Multistage progressive learning based speech enhancement using time-frequency attentive squeezed temporal convolutional networks

Speech enhancement (SE) is an important method for improving speech quality and intelligibility in noisy environments. An effective speech enhancement model depends on precise modelling of the long-range dependencies of noisy speech. Several recent studies have examined ways to enhance speech by capturing the long-term contextual information. For speech enhancement, the time-frequency (T-F) distribution of speech spectral components is also important, but is usually ignored in these studies. The multi-stage learning method is an effective way to integrate various deep-learning modules at the same time. The benefit of multi-stage training is that the optimization target can be iteratively updated stage by stage. In this paper speech enhancement is investigated by multi-stage learning using a multistage structure in which time-frequency attention (TFA) blocks are followed by stacks of squeezed temporal convolutional networks (S-TCN) with exponentially increasing dilation rates. To reinject original information into later stages, a feature fusion (FF) block is inserted at the input of later stages to reduce the possibility of speech information being lost in the early stages. The S-TCN blocks are responsible for temporal sequence modelling task. The time-frequency attention (TFA) is a simple but effective network module that explicitly exploits position information to generate a 2D attention map to characterise the salient T-F distribution of speech by using two branches, time-frame attention and frequency attention in parallel. A set of utterances from the LibriSpeech and Voicebank databases are used to evaluate the performance of the proposed SE. Extensive experiments have demonstrated that the proposed model consistently improves the performance over existing baselines across two widely used objective metrics such as PESQ and STOI. The average PESQ and STOI for proposed model are boosted by a factor of 41.7% and 5.4% for Libri speech dataset, 36.10% and 3.1% for Voice bank dataset as compared to noisy speech. Additionally, we explored the generalization of the proposed TFA-S-TCN model across different speech datasets through cross data base analysis. A significant improvement in system robustness to noise is also shown by our evaluation results using the TFA module.

Pinging the brain to reveal the hidden attentional priority map using encephalography

Attention has been usefully thought of as organized in priority maps – putative maps of space where attentional priority is weighted across spatial regions in a winner-take-all competition for attentional deployment. Recent work has highlighted the influence of past experiences on the weighting of spatial priority – called selection history. Aside from being distinct from more well-studied, top-down forms of attentional enhancement, little is known about the neural substrates of history-mediated attentional priority. Using a task known to induce statistical learning of target distributions, in an EEG study we demonstrate that this otherwise invisible, latent attentional priority map can be visualized during the intertrial period using a ‘pinging’ technique in conjunction with multivariate pattern analyses. Our findings not only offer a method of visualizing the history-mediated attentional priority map, but also shed light on the underlying mechanisms allowing our past experiences to influence future behavior.

Cyclic Generative Attention-Adversarial Network for Low-Light Image Enhancement.

Images captured under complex conditions frequently have low quality, and image performance obtained under low-light conditions is poor and does not satisfy subsequent engineering processing. The goal of low-light image enhancement is to restore low-light images to normal illumination levels. Although many methods have emerged in this field, they are inadequate for dealing with noise, color deviation, and exposure issues. To address these issues, we present CGAAN, a new unsupervised generative adversarial network that combines a new attention module and a new normalization function based on cycle generative adversarial networks and employs a global-local discriminator trained with unpaired low-light and normal-light images and stylized region loss. Our attention generates feature maps via global and average pooling, and the weights of different feature maps are calculated by multiplying learnable parameters and feature maps in the appropriate order. These weights indicate the significance of corresponding features. Specifically, our attention is a feature map attention mechanism that improves the network's feature-extraction ability by distinguishing the normal light domain from the low-light domain to obtain an attention map to solve the color bias and exposure problems. The style region loss guides the network to more effectively eliminate the effects of noise. The new normalization function we present preserves more semantic information while normalizing the image, which can guide the model to recover more details and improve image quality even further. The experimental results demonstrate that the proposed method can produce good results that are useful for practical applications.

Spatial attention-based residual network for human burn identification and classification

Diagnosing burns in humans has become critical, as early identification can save lives. The manual process of burn diagnosis is time-consuming and complex, even for experienced doctors. Machine learning (ML) and deep convolutional neural network (CNN) models have emerged as the standard for medical image diagnosis. The ML-based approach typically requires handcrafted features for training, which may result in suboptimal performance. Conversely, DL-based methods automatically extract features, but designing a robust model is challenging. Additionally, shallow DL methods lack long-range feature dependency, decreasing efficiency in various applications. We implemented several deep CNN models, ResNeXt, VGG16, and AlexNet, for human burn diagnosis. The results obtained from these models were found to be less reliable since shallow deep CNN models need improved attention modules to preserve the feature dependencies. Therefore, in the proposed study, the feature map is divided into several categories, and the channel dependencies between any two channel mappings within a given class are highlighted. A spatial attention map is built by considering the links between features and their locations. Our attention-based model BuRnGANeXt50 kernel and convolutional layers are also optimized for human burn diagnosis. The earlier study classified the burn based on depth of graft and non-graft. We first classified the burn based on the degree. Subsequently, it is classified into graft and non-graft. Furthermore, the proposed model performance is evaluated on Burns_BIP_US_database. The sensitivity of the BuRnGANeXt50 is 97.22% and 99.14%, respectively, for classifying burns based on degree and depth. This model may be used for quick screening of burn patients and can be executed in the cloud or on a local machine. The code of the proposed method can be accessed at https://github.com/dhirujis02/Journal.git for the sake of reproducibility.