Search

Loop closure detection is a key component of visual simultaneous localization and mapping (VSLM), which can effectively reduce the cumulative error of the system and improve the accuracy of mapping. The current loop closure detection method using deep learning can obtain accurate scene descriptions, but it is difficult to cope with the challenge of scene changes. This paper obtains complementary feature maps by fusing the shallow and deep image features of the ResNet50 network, and we incorporate a multiscale channel attention mechanism and a spatial attention mechanism after each layer of the ResNet50 network. The model can effectively extract discriminative scene landmarks, suppress the effects of irrelevant local features on similarity and be more robust to the problem of scene change. The method in this paper has been tested on several publicly available datasets and compared with mainstream methods. The experimental results show that the proposed method significantly outperforms mainstream methods in terms of accuracy-recall performance. [ABSTRACT FROM AUTHOR]

Background: There are many conflicting findings on the gaze cueing effect (GCE) of emotional facial expressions. This study aimed to investigate whether an averted gaze, accompanied by a fearful expression of different durations, could enhance attentional orientation, as measured by a participant's eye movements. Methods: Twelve participants (3 females) completed the gaze cue task, reacting to a target location after observing changes in the gaze and expression of a face illustrated on a computer screen. Meanwhile, participants' eye movements were monitored by electrooculography. The GCE was calculated by reaction time as an indicator of attention shift. Results: The analysis of the overall data did not find a significant effect of fearful facial expressions on the GCE. However, analysis of trial data that excluded a participant's eye movement data showed that brief (0, 100 ms) presentation of the fearful facial expression enhanced the GCE compared to that during a neutral facial expression, although when the presentation time of the fearful expression was increased to 200 or 400 ms, the GCE of the fearful expression was at the same level as when model showed a neutral expression. Conclusions: The results suggest that the attention-enhancing effect of gaze cues induced by rapidly presented fearful expressions occurs only when the effect of eye movement trials is excluded. This effect may be mediated by reflexively neural circuits in the amygdala that process threatening stimuli. However, as the expression duration increased, the fearful expression's attention-enhancing effect decreased. We suggest that future studies on the emotion modulation of GCE should consider the negative effects of participants' saccades and blinks on the experimental results. [ABSTRACT FROM AUTHOR]

Blood vessel segmentation in fundus images is necessary for the diagnosis of ophthalmic diseases. In recent years, deep learning has achieved eminent performance in blood vessel segmentation, and there still exist challenges to reduce misidentification and improve microvascular segmentation accuracy. One reason is that traditional Convolutional Neural Network (CNN) can not effectively extract multiscale information and discard the unnecessary information. Another reason is we can't explain why some blood vessels fail to be identified. On the one hand, this paper proposes a Dual Attention Res2UNet (DA-Res2UNet) model. The DA-Res2UNet model uses Res2block rather than CNN to obtain more multiscale information and adds Dual Attention to help the model focus on important information and discard unnecessary information. On the other hand, the explainable method based on a pre-trained fundus image generator is adopted to explore how the model identifies blood vessels. We deduce several special situations that lead to the misidentification based on the model's explanation and adjust the dataset for these special cases. The adjusted datasets significantly reduce the misidentification in the CHASE_DB1 dataset. Finally, the model trained by the adjusted datasets achieves the state-of-the-art F1-score of 81.88%, 82.77%, and 83.96% on the CHASE_DB1, DRIVE and STARE datasets, respectively. [ABSTRACT FROM AUTHOR]

Chest radiographs, among other medical imaging, are the most significant and effective diagnostic tools for detecting lung disorders. Numerous research is being done to develop reliable and automatic diagnostic systems for detecting diseases using chest radiographs. Pneumothorax is a potentially fatal condition that needs early diagnosis and treatment. Artificial Intelligence (AI) approaches have offered promising results in medical imaging. Different AI-based approaches for classifying pneumothorax using medical images have been proposed. However, there is limited medical imaging available for the identification of pneumothorax. This work aims to develop a model to detect pneumothorax in chest X-ray images by combining xception network with an attention module. The proposed model was experimented on 2,597 chest X-ray images and has achieved training accuracy of 99.18%, validation accuracy of 87.53% and average AUC (Area under the ROC Curve) of 90.00%. [ABSTRACT FROM AUTHOR]

Image inpainting based on deep learning has been greatly improved. The original purpose of image inpainting was to repair some broken photos, such as inpainting artifacts. However, it may also be used for malicious operations, such as destroying evidence. Therefore, detection and localization of image inpainting operations are essential. Recent research shows that high-pass filtering full convolutional network (HPFCN) is applied to image inpainting detection and achieves good results. However, those methods did not consider the spatial location and channel information of the feature map. To solve these shortcomings, we introduce the squeezed excitation blocks (SE) and propose a high-pass filter attention full convolutional network (HPACN). In feature extraction, we apply concurrent spatial and channel attention (scSE) to enhance feature extraction and obtain more information. Channel attention (cSE) is introduced in upsampling to enhance detection and localization. The experimental results show that the proposed method can achieve improvement on ImageNet. [ABSTRACT FROM AUTHOR]

The abdominal tumor is a general term for tumors in kidney, liver and pancreas. Accurate segmentation of abdominal tumors is essential for their treatment. However, the varying shapes and sizes of abdominal organs result in significant differences in tumor regions. Existing convolution neural networks (CNNs) can only accurately segment individual abdominal tumors, lacking sufficient generalizability. We aim to design a network that can achieve good segmentation results for different abdominal tumors. To this end, a Spider-net to segment tumors is presented in this paper, which consists of a high-resolution multi-scale attention encoder and a full-attention decoder. Additionally, scale attention that integrates channel attention and spatial attention is designed for generating output. We have also designed a classification branch to distinguish whether the segmented region is a real tumor area or another benign lesion. We train and evaluate the Spider-net on three different organs: the kidney, pancreas, and liver. Spider-net achieves state-of-the-art results compared to methods that only use CNNs or transformers. Code are available at https://github.com/h2440222798/HRMA. [Display omitted] • Convolution and vision transformer are combined in a high-resolution structure. • A lightweight patch embedding and cross-layer fusion are employed for reducing the parameters of the encoder. • The fusions of different attention modules are proposed for decoding. • Incorporating both channel and spatial attention into the scale attention for generating the output. • A new classification branch are proposed for classify if current slice contains tumor. [ABSTRACT FROM AUTHOR]

The debate about the scope and limits of unconscious visual processing under continuous flash suppression (CFS) has created a heterogeneous set of divergent findings that are yet to be reconciled. Attention has been suggested as an important factor in modulating the processing of suppressed visual information under CFS. Specifically, Eo et al. (2016) reported that semantic processing under CFS can be significantly facilitated when spatial attention is diverted away from the suppressed stimulus. Based on event-related potential (ERP) findings involving the N400, they proposed that inattention attenuates interocular suppression and thereby makes semantic processing available unconsciously, potentially reconciling conflicting evidence in the literature. In this study, we aimed to further investigate the "CFS-attenuation-by-inattention" hypothesis using functional magnetic resonance imaging (fMRI) and multivariate pattern analysis (MVPA). We tested whether the decodability of object category increases under CFS when attention is diverted away from the suppressed stimulus in a spatial cueing task. Our results provide no evidence for the "CFS-attenuation-by-inattention" hypothesis, but show higher decoding accuracies for visible stimuli than for invisible stimuli. We discuss the implications of our findings for the important endeavor of trying to reconcile the divergent reports of unconscious processing under CFS. [ABSTRACT FROM AUTHOR]

Facial expression recognition (FER) is to separate the specific expression state from the given static image or video to determine the psychological emotions of the recognized object, the realization of the computer's understanding and recognition of facial expressions have fundamentally changed the relationship between human and computer, to achieve better human computer interaction (HCI). In recent years, FER has attracted widespread attention in the fields of HCI, security, communications and driving, and has become one of the research hotspots. In the mobile Internet era, the need for lightweight networking and real-time performance is growing. In this paper, a lightweight A-MobileNet model is proposed. First, the attention module is introduced into the MobileNetV1 model to enhance the local feature extraction of facial expressions. Then, the center loss and softmax loss are combined to optimize the model parameters to reduce intra-class distance and increase inter-class distance. Compared with the original MobileNet series models, our method significantly improves recognition accuracy without increasing the number of model parameters. Compared with others, A- MobileNet model achieves better results on the FERPlus and RAF-DB datasets. [ABSTRACT FROM AUTHOR]

Target representation is crucial for visual tracking. Most Siamese-based trackers try their best to establish target models by using various deep networks. However, they neglect the exploration of correlation among features, which leads to the inability to learn more representative features. In this paper, we propose a spatial attention inference model for cascaded Siamese tracking with dynamic residual update strategy. First, a spatial attention inference model is constructed. The model fuses interlayer multi-scale features generated by dilation convolution to enhance the spatial representation ability of features. On this basis, we use self-attention to capture interaction between target and context, and use cross-attention to aggregate interdependencies between target and background. The model infers potential feature information by exploiting the correlations among features for building better appearance models. Second, a cascaded localization-aware network is introduced to bridge a gap between classification and regression. We propose an alignment-aware branch to resample and learn object-aware features from the predicted bounding boxes for obtaining localization confidence, which is used to correct the classification confidence by weighted integration. This cascaded strategy alleviates the misalignment problem between classification and regression. Finally, a dynamic residual update strategy is proposed. This strategy utilizes the Context Fusion Network (CFNet) to fuse the templates of historical and current frames to generate the optimal templates. Meanwhile, we use a dynamic threshold function to determine when to update by judging the tracking results. The strategy uses temporal context to fully explore the intrinsic properties of the target, which enhances the adaptability to changes in the target's appearance. We conducted extensive experiments on seven tracking benchmarks, including OTB100, UAV123, TC128, VOT2016, VOT2018, GOT10k and LaSOT, to validate the effectiveness of our proposed algorithm. • Innovative spatial attention model boosts spatial feature representation. • Target-background interdependencies captured via self & cross-attention. • Cascaded localization-aware network solves classification-regression misalignment. • Dynamic residual strategy with CFNet adapts to target appearance changes. • Extensive validation shows superior performance on diverse tracking benchmarks. [ABSTRACT FROM AUTHOR]

Cervical cancer is a common gynecological tumor treated predominantly with radiotherapy for locally advanced cases. However, despite treatment, almost one-third of patients experience recurrence within 18 months. Accurate prediction of patient response to therapy is critical for selecting optimal treatment plans. Currently, MRI images are manually segmented to identify the tumor region and predict treatment response using image information within the tumor. However, current methods treat segmentation and response prediction as separate tasks and manual segmentation can be expensive. To address these issues, we propose a spatial and task attention network that simultaneously segments the tumor and predicts the response to cervical cancer radiotherapy. Our approach employs a spatial attention module to focus on the tumor region and a task attention module to explore the correlation between tumor segmentation and treatment response prediction, achieving automatic segmentation of the tumor. We retrospectively collected MRI images from 138 patients with locally advanced cervical cancer before radiotherapy and conducted 5-fold cross-validation experiments, demonstrating that our method achieves competitive results. [Display omitted] • We devise a multitask architecture that simultaneously predicts the treatment response and segments the tumor region for locally advanced cervical cancer radiotherapy. • The proposed spatial attention module can allow the network to focus more on the tumor region. The attention module can effectively improve the performance of both tasks by utilizing the inherent correlation of them. • Extensive experiments using retrospective locally advanced cervical cancer data showcase the effectiveness of our proposed method in predicting treatment response and segmenting tumors. [ABSTRACT FROM AUTHOR]

Background Spatial attention relatively increases the power of neural 10-Hz alpha oscillations in the hemisphere ipsilateral to attention, and decreases alpha power in the contralateral hemisphere. For gamma oscillations (>40 Hz), the opposite effect has been observed. The functional roles of lateralised oscillations for attention are currently unclear. Hypothesis If lateralised oscillations are functionally relevant for attention, transcranial stimulation of alpha versus gamma oscillations in one hemisphere should differentially modulate the accuracy of spatial attention to the ipsi-versus contralateral side. Methods 20 human participants performed a dichotic listening task under continuous transcranial alternating current stimulation (tACS, vs sham) at alpha (10 Hz) or gamma (47 Hz) frequency. On each trial, participants attended to four spoken numbers on the left or right ear, while ignoring numbers on the other ear. In order to stimulate a left temporo-parietal cortex region, which is known to show marked modulations of alpha power during auditory spatial attention, tACS (1 mA peak-to-peak amplitude) was applied at electrode positions TP7 and FC5 over the left hemisphere. Results As predicted, unihemispheric alpha-tACS relatively decreased the recall of targets contralateral to stimulation, but increased recall of ipsilateral targets. Importantly, this spatial pattern of results was reversed for gamma-tACS. Conclusions Results provide a proof of concept that transcranially stimulated oscillations can enhance spatial attention and facilitate attentional selection of speech. Furthermore, opposite effects of alpha versus gamma stimulation support the view that states of high alpha are incommensurate with active neural processing as reflected by states of high gamma. [ABSTRACT FROM AUTHOR]

The main idea of a traditional Steady State Visually Evoked Potentials (SSVEP)-BCI is the activation of commands through gaze control. However, the widely named “dependent” SSVEP-BCIs might not be applicable for patients with ocular motor impairments or severe neuromuscular problems. Nevertheless, an “independent” SSVEP-BCIs might be a potential approach to solve this problem. This study presents a novel independent-BCI based on SSVEP using Figure-Ground Perception (FGP), terminology widely known and used in Gestalt psychology for object recognition by means of changes in perception. This BCI proposes to identify two different targets that represent commands in a limited visual space without needing to shift the gaze by the paradigm of covert attention. For that purpose, the well-known example of Rubin's face-vase in FGP was used. The traditional EEG signal analysis consists of three steps: filtering, feature extraction and classification. In this work, two techniques were used for performance comparison, and the classification was obtained through a criterion of maxima for both techniques. Ten subjects participated in this study in offline tests and five subjects for online tests. The flickering frequencies were 15.0 Hz (vase) and 11.0 Hz (faces). Our results demonstrate that the electrode Oz is the best channel for characterization of visual perception, from a quantitative point of view based on the canonical correlation, after a channel analysis by independent way. Regarding the classification, MSI technique was more accurate in relation to CCA, in all the cases with same conditions, either using three electrodes or a single electrode (Oz), even for different window lengths. The online performance appeared to decrease as participants switched from Face (82.7%) to Vase (76%) stimulus. These results are consistent with our results in offline tasks. Muscular activity related to the eye movements was also evaluated using a commercial device of eye tracking (Eye Tribe). These findings strongly support the hypothesis of visual selectivity by means of perception and neural mechanism of spatial attention. [ABSTRACT FROM AUTHOR]

Thirteen international netballers viewed computerized static images of scenarios taken from netball open play. Two 'team mates', each marked by one opponent, could be seen in each image; each team mate-opponent pair was located on opposite sides of the display, such that a binary response was required ('left' or 'right') from the participant, in order to select a team mate to whom they would pass the ball. For each trial, a spoken word ("left"/ "right") was presented monaurally at the onset of the visual image. Spatially invalid auditory cues (i.e., in the ear contralateral to the correct passing option [as judged by three netball experts]), reduced performance accuracy relative to valid ones. Semantically invalid cues (e.g., a call of "left" when the target was right-located), increased response times relative to valid ones. However, there were no accompanying changes in visual attention to the team mates and their markers. The effects of auditory cues on covert attentional shifts and decision-making are discussed. [ABSTRACT FROM AUTHOR]

Introduction: Visual searching is an essential component of many everyday activities. Search training is practised as part of occupational therapy to improve performance skills both in people with hemianopia and those with spatial inattention post stroke. Evaluation of the effectiveness of such training first requires a systematic and detailed description of the intervention. To this end, this study describes the practice of a specialist occupational therapist. Method: Single sessions of intervention delivered by the occupational therapist to five participants with visual search disorders post stroke were video recorded. The recordings were analysed for content using a framework approach. Results: The occupational therapist educated participants about the impact of their visual impairment on everyday activities. She used instructions, spatial cueing, placement of objects and verbal feedback to train increased amplitudes of eye and head movements, to direct attention into the blind field or neglected space and to encourage systematic searching during occupations and activities. Activities were graded by manipulating the area for attention and complexity in the environment. Conclusion: This investigation provides a detailed description of a specialist occupational therapist's community-based intervention for improving visual search post stroke. [ABSTRACT FROM AUTHOR]

• Develop a CNN model using color fundus photographs to recognize the multi-label ophthalmological disease. • The model produces eight values indicating whether the input is healthy or associated with seven categories of anomalies. • A spatial attention module is incorporated into the model to improve the network feature extraction capability. • The proposed model significantly improves recognition performance using a publicly accessible CFP dataset by refining the multi-label model structure and loss function. • Analyze the error classification approaches and provide error correction strategies that enable the model to self-correct, increasing its robustness. Color fundus photographs enable the observation of numerous critical biomarkers and early-onset lesions associated with illnesses. Due to its non-invasive and cost-effective nature, this approach can be used for large-scale screening of fundus disorders. In recent years, most applications of data-driven deep learning approaches to fundus illnesses have been based on color fundus photographs. However, current screening methods for fundus illnesses are mainly focused on identifying particular diseases. The majority of published models for multi-disease recognition are ensembles of several binary classification networks, and an ensemble network consumes more resources than a single network. Additionally, many fundus illness recognition models make direct use of established networks for natural picture processing without performing structural optimization, which may reduce disease classification accuracy. To address these issues, we optimized the network structure to build a single multi-label fundus disease recognition model. Specifically, we began by selecting EfficientNet-B4 as a backbone network, then modified the output layer to create a multi-label recognition model. Based on this model, the spatial attention structure was extended to improve the network's feature extraction capability. The focal loss was used to increase classification accuracy by mitigating the training effects of data imbalance. Our proposed model significantly improves recognition performance on a publicly available dataset compared to other baseline networks. When spatial attention and focal loss function are introduced into the backbone model, the baseline values of all relevant evaluation indicators are improved without significant increase in computational cost. Additionally, this article provides two error correction strategies for multi-label classification issues: mutual exclusion and super label space. When a more robust error correction strategy is employed with the model, F1 increases by 2.86% in comparison to the original performance of EfficientNet-B4. [ABSTRACT FROM AUTHOR]

Motivational accounts of pain behavior and disability suggest that persisting attempts to avoid or control pain may paradoxically result in heightened attention to pain-related information. We investigated whether attempts to control pain prioritized attention to the location where pain was expected, using a tactile change detection paradigm. Thirty-seven undergraduate students had to detect changes between 2 consecutively presented patterns of tactile stimuli at various body locations. One of the locations was made threatening by occasionally administering a pain-eliciting stimulus. Half of the participants (pain control group) were encouraged to actively avoid the administering of pain by pressing a button as quickly as possible, whereas the other participants (comparison group) were not. The actual amount of painful stimuli was the same in both groups. Results showed that in the comparison group, the anticipation of pain resulted in better detection of tactile changes at the pain location than at the other locations, indicating an attentional bias for the threatened location. Crucially, the pain control group showed a similar attentional bias, but also when there was no actual presence of threat. This suggests that although threat briefly prioritized the threatened location, the goal to control pain did so in a broader, more context-driven manner. [ABSTRACT FROM AUTHOR]

• A dual attention mechanism is proposed which can highlight the meaningful local patterns contained in the skin lesion regions, enhancing the feature representation and the interpretability of the proposed network at the same time. • An auxiliary learning mechanism which contains auxiliary supervision and KL divergence based regularization is proposed. The KL regularization can make the two auxiliary supervision branches collaborate with each other during training through mutual knowledge transferring, and guide the network to extract the meaningful local pattern features contained in the skin lesion region in a weakly supervised manner. Besides, it brings in strong regularization which makes our proposed network avoid over-fitting when training on the small training data. • The proposed network gained the state-of-the-art performance for skin lesion classification, regardless of binary- or multi-classification. Besides, the robustness and interpretability of the proposed network are strong which can promote its clinical application. Skin lesion varies greatly in appearance, and its classification task suffers from large inter-class similarity and intra-class variation, thus the subtle differences of local pattern contained in the skin lesion regions are critical for its classification. In this paper, we propose a dual attention based network for skin lesion classification with auxiliary learning. The dual attention mechanism includes the spatial attention (SA) and the channel attention (CA) modules. The SA module can focus on the skin lesion region feature with reduced irrelevant artifacts features. In the subsequent CA module, it first captures the non-local based global feature of the skin lesion region and then generates the feature channels reweighting vector, which is used to further refine the meaningful local pattern feature contained in the skin lesion region. Therefore, the performance and the interpretability of the proposed network are enhanced at the same time. The proposed auxiliary learning contains two auxiliary supervision branches and KL regularization. The KL regularization makes the two auxiliary supervision branches collaborate with each other during training through mutual knowledge transferring. The introduced strong regularization can guide the dual attention mechanism to focus on the meaningful local pattern features in a weakly supervised manner and make the network avoid overfitting on small training data. Without extra training data, our proposed network can outperform current competition winners on several datasets, regardless of binary- or multi-classification. The proposed network is robust enough and owns strong interpretability which promotes its clinical application. [ABSTRACT FROM AUTHOR]

Abstract: Cognitive factors like attention can modulate the brain activities in different cortical areas. The brain activities can be measured using different systems with different spatial and temporal resolutions. The magnetoencephalography (MEG) is one of those systems that can measure the brain activities in a high temporal resolution. Here the brain signals have been recorded using the MEG system from different brain areas of human subjects while doing a visual spatial attention task. These signals have been forwarded to a pattern recognition system for the possibility of predicting the attentional state of the subjects in two different positions. The proposed hybrid system consists of channel selection using Bayesian approach, feature extraction using the wavelet packet and feature selection based on entropy-based method. The final classifier was selected to be Naive Bayesian classifier for attentional state prediction. The results indicate that the proposed system can predict the location of the attended stimulus with a high accuracy, so it can be helpful for brain–computer interface (BCI) applications. [Copyright &y& Elsevier]

Abstract: Background: The right posterior parietal cortex (rPPC) and the right frontal eye field (rFEF) form part of a network of brain areas involved in orienting spatial attention. Previous studies using transcranial magnetic stimulation (TMS) have demonstrated that both areas are critically involved in the processing of conjunction visual search tasks, since stimulation of these sites disrupts performance. Objective: This study investigated the effects of long term neuronal modulation to rPPC and rFEF using transcranial direct current stimulation (tDCS) with the aim of uncovering sharing of these resources in the processing of conjunction visual search tasks. Methods: Participants completed four blocks of conjunction search trials over the course of 45 min. Following the first block they received 15 min of either cathodal or anodal stimulation to rPPC or rFEF, or sham stimulation. Results: A significant interaction between block and stimulation condition was found, indicating that tDCS caused different effects according to the site (rPPC or rFEF) and type of stimulation (cathodal, anodal, or sham). Practice resulted in a significant reduction in reaction time across the four blocks in all conditions except when cathodal tDCS was applied to rPPC. Conclusions: The effects of cathodal tDCS over rPPC are subtler than those seen with TMS, and no effect of tDCS was evident at rFEF. This suggests that rFEF has a more transient role than rPPC in the processing of conjunction visual search and is robust to longer-term methods of neuro-disruption. Our results may be explained within the framework of functional connectivity between these, and other, areas. [Copyright &y& Elsevier]

Spatial visual attention enhances information processing within its focus. Vision at an attended location is faster, more accurate, of higher spatial resolution, and has an enhanced sensitivity for fine changes. Earlier hypotheses suggest that the neuronal mechanisms of these processes are based on the interactions among different neuronal groups by means of cortical oscillations in the gamma range. The aim of the current study was to modulate these oscillations externally, using a new technique called transcranial alternating current stimulation (tACS). We investigated the effect of covert spatial attention within and outside its focus by probing contrast sensitivity and contrast discrimination at high resolution across the visual field of 20 healthy human subjects. While applying 40, 60, and 80 Hz tAC stimulation over the primary visual cortex (V1), subjects’ contrast-discrimination thresholds were obtained using two different conditions: in the first condition we presented a black disc as a peripheral cue that automatically attracted the subject’s attention, whereas there was no cue in the second condition. We found that the spatial profile of contrast sensitivity was not affected by the stimulation. Contrast-discrimination thresholds on the other hand decreased significantly during 60 Hz tACS, whereas there was no effect of 40 and 80 Hz stimulation. These results suggest that attention plays an important role in contrast discrimination based on V1 activities that are influences by gamma range tACS stimulation. [ABSTRACT FROM AUTHOR]

Copyright of Psychologie Française is the property of Elsevier B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

We investigated whether words relevant to a person's current goal and words related to that goal influence the orienting of attention even when an intention to attend to the goal-relevant and goal-related stimuli is not present. Participants performed a modified spatial cueing paradigm combined with a second task that induced a goal. The results showed that the induced goal led to the orienting of attention to goal-relevant words in the spatial cueing task. This effect was not found for goal-related words. The results provide evidence for accounts of automatic goal pursuit, which state that goals automatically guide attention to goal-relevant events. [ABSTRACT FROM AUTHOR]

Non-uniform blind motion deblurring is a challenging yet important problem in image processing that receives enduring attention in the last decade. The non-uniformity nature of motion blurring leads to great variations on the blurring effects across image regions and over different images, which makes it very difficult to train an end-to-end deblurring neural network (NN) with good generalization performance. This paper introduces an attention mechanism for the blind deblurring NN, including both spatial and channel attention, so as to effectively handle the significant spatial variations on blurring effects. In the attention mechanism, the spatial attention is introduced in both the encoder for discriminative exploitation of image edges and smooth regions and the decoder for discriminative treatment on different regions with different blurring effects. The channel attention is introduced for better generalization performance of the NN, as it allows adaptive weighting on intermediate features for a particular image. Building such an attention mechanism into a multi-scale encoder–decoder framework, an attentive NN is developed for practical non-uniform blind image deblurring. The experiments on several benchmark datasets show that the proposed NN can effectively restore the images degraded by spatially-varying blurring, with state-of-the-art performance. • A joint spatial-channel attention-based DNN for image deblurring. • Joint attention improves generalizability and performance. • We show different roles of spatial attention played in encoder and decoder. • SOTA performance achieved. [ABSTRACT FROM AUTHOR]

Recently, deep learning based inpainting methods have shown promising performance, in which some multi-scale networks are introduced to characterize image content in both details and structures. However, few of these networks explore local spatial components under different receptive fields and internal connection between multi-scale feature maps. In this paper, we propose a novel multi-scale attention network (MSA-Net) to fill the irregular missing regions, in which a multi-scale attention group (MSAG) with several multi-scale attention units (MSAUs) is introduced for fully analysing the features from shallow details to high-level semantics. In each MSAU, an attention based spatial pyramid structure is designed to capture the deep features from different receptive fields. In this network, attention mechanisms are explored for boosting the representation power of MSAU, where spatial attention is combined with each scale to highlight the most probably attentive spatial components and the channel attention is used as a globally semantic detector to build the connection between the multiple scales. Furthermore, for better inpainting results, a max pooling based mask update method is utilized to predict the missing parts from the border regions to the inside. Finally, experiments on Places2 dataset and CelebA dataset demonstrate that the proposed method can achieve better results than the previous inpainting methods. • A multi-scale attention network is proposed for image inpainting • An attention based spatial pyramid structure is designed for multi-scale features • A mask update method is used to fill the holes from the border regions to the inside [ABSTRACT FROM AUTHOR]

• The conditional entropy technique enabled simultaneous analysis of heart-brain rhythms. • Directional coupling information C (heart→brain) and C (brain→heart) differentiated internally and externally operative attention. • Attention states were examined using simultaneous recording of EEG and ECG signals. • A modified Posner's spatial orienting task used to assess A I and A E attention. • Internal and external attention states investigated using identical stimulus design. Heart and brain interactions mediate human cognition. This investigation identifies heart-brain interactions during internally operative attention (A I) and externally operative attention (A E). A I attention involves short term memory, whereas A E attention deals with automatic and transient response to objects in the external world. A modified Posner's spatial orienting task used to differentiate A I and A E attention. Heart and brain rhythms recorded in fourteen healthy participants. Functional coupling from heart-to-brain (C h e a r t → b r a i n) and brain-to-heart (C b r a i n → h e a r t) time series derived using an information domain approach based on conditional entropy. The experimental results showed that low-frequency power of heart rate variability (HRV-LF) and sympathovagal balance (LF/HF ratio) during A E significantly increased compared with that for A I. Furthermore, the information flow from heart-to-brain increased and decreased form brain-to-heart during A E as compared to A I. Also, opposite trend in relationship noted between coupling index (C i → j) and HRV-LF during A I and A E attention. The conditional entropy technique enabled simultaneous analysis of heart-brain rhythms to identify heart-brain interactions during A I and A E attention. [ABSTRACT FROM AUTHOR]