Dimension-aware attention for efficient mobile networks

Mo, Rongyun; Lai, Shupeng; Yan, Yan; Chai, Zhenhua; Wei, Xiaolin

doi:10.1016/j.patcog.2022.108899

Cited by 4 publications

(2 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This helps to produce three feature matrices based on the dimension. Dimension-wise convolutions are used to encode spatial and channel-wise information in the input tensor efficiently [46,47]. The input tensor M has dimensions w, h, and d, where w represents width, d represents depth, and h represents height.…”

Section: ) Swin Transformer Trackmentioning

confidence: 99%

A Dimension Centric Proximate Attention Network and Swin Transformer for Age-Based Classification of Mild Cognitive Impairment From Brain MRI

Illakiya,

Karthik

2023

IEEE Access

View full text Add to dashboard Cite

The early identification and treatment of Mild Cognitive Impairment (MCI) play a crucial role in managing the risk of Alzheimer's disease (AD). However, current methods for categorizing progressive MCI and stable MCI based on brain MRI scans have proven insufficient due to the subtle nature of the features involved. This research aims to improve the effectiveness of MCI classification through the utilization of a Deep Learning (DL) network. The primary objective of this work is to improve the feature representation of brain MRI scans for more accurate classification. The proposed model is a hybrid MCI classification system that integrates three components: the Swin Transformer, the Dimension Centric Proximity Aware Attention Network (DCPAN), and the Age Deviation Factor (ADF). The proposed network achieves better classification results through a unique feature fusion approach that combines global, local, proximal features, and dimensional dependencies. It effectively combines fine-grained details with broader contextual information to extract discriminative features. Experimental results demonstrate the effectiveness of the proposed network, achieving an accuracy of 79.8%, precision of 76.6%, recall of 80.2%, and an F1-score of 78.4% when evaluated on the ADNI dataset.

show abstract

Section: ) Swin Transformer Trackmentioning

confidence: 99%

A Dimension Centric Proximate Attention Network and Swin Transformer for Age-Based Classification of Mild Cognitive Impairment From Brain MRI

Illakiya,

Karthik

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Existing attention mechanisms, such as CBAM and SE, typically employ global maximum pooling or global average pooling operations, which can result in the loss of spatial information. In contrast, the Coordinate Attention (CA) module [34] incorporates location information into channel attention, allowing for the consideration of both channel and location information. This integration effectively increases the emphasis on the target to be recognized within the image.…”

Section: Coordinate Attentionmentioning

confidence: 99%

Tunnel Lining Multi-Defect Detection Based on an Improved You Only Look Once Version 7 Algorithm

Juan,

Long-Xi,

Hui-Ping

2023

IEEE Access

View full text Add to dashboard Cite

In the domain of tunnel lining defect detection, object detection algorithms have been widely employed. However, existing algorithms suffer from inadequate extraction of global information and low detection accuracy. To address these issues, a novel algorithm called Tunnel Defect Detection You Only Look Once (TDD-YOLO) is proposed, leveraging the YOLOv7 framework. The TDD-YOLO algorithm incorporates several enhancements to improve global and local information extraction capabilities, thereby enhancing defect detection accuracy. Firstly, MobileViT is utilized as the backbone feature extraction network, augmenting the network's ability to extract comprehensive information from both global and local contexts. Secondly, a Coordinate Attention (CA) module is introduced after the upsampling and downsampling stages of the feature pyramid network. This module highlights defect-related features while eliminating background interference. Lastly, a convolutional module called TP Block is devised to further enhance the network's feature extraction capability with reduced computational complexity. To validate the effectiveness of the proposed algorithm, a comparative analysis is conducted against five existing algorithms: SSD, Faster-RCNN, EfficientDet, YOLOv5, and YOLOv7. Experimental results demonstrate that the TDD-YOLO algorithm achieves superior performance with an F1 score of 77.43% and a mean Average Precision (mAP) of 77.52%. These results surpass those of the other five algorithms, establishing the TDD-YOLO algorithm as the most accurate and suitable solution for defect detection tasks in tunnels.

show abstract