Self-Attention Multilayer Feature Fusion Based on Long Connection
Feature fusion is an important part of building high-precision convolutional neural networks. In the field of image classification, though widely used in processing multiscale features of the same layer and short connections in the same receptive field, feature fusion is rarely used in long connecti...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2022-01-01
|
| Series: | Advances in Multimedia |
| Online Access: | http://dx.doi.org/10.1155/2022/9973814 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832553480521252864 |
|---|---|
| author | Chu Yuezhong Wang Jiaqing Liu Heng |
| author_facet | Chu Yuezhong Wang Jiaqing Liu Heng |
| author_sort | Chu Yuezhong |
| collection | DOAJ |
| description | Feature fusion is an important part of building high-precision convolutional neural networks. In the field of image classification, though widely used in processing multiscale features of the same layer and short connections in the same receptive field, feature fusion is rarely used in long connection operations across receptive fields. In order to fuse the high- and low-level features of image classification, a feature fusion module SCFF (selective cross-layer feature fusion) for long connections is designed in this work. The SCFF can connect the long-distance feature maps in different receptive fields in a top-down order and apply the self-attention mechanism to fuse them two by two. The final fusion result is used as the input of the classifier. In order to verify the effectiveness of the model, the image classification experiment was done on a number of typical datasets. The experimental results prove that the model can fit the existing convolutional neural network well and effectively improve the classification accuracy of the convolutional network only at the cost of a small amount of calculation. |
| format | Article |
| id | doaj-art-863bf3f92c7f43b3834c1f381b0c9c52 |
| institution | Kabale University |
| issn | 1687-5699 |
| language | English |
| publishDate | 2022-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Advances in Multimedia |
| spelling | doaj-art-863bf3f92c7f43b3834c1f381b0c9c522025-02-03T05:53:50ZengWileyAdvances in Multimedia1687-56992022-01-01202210.1155/2022/9973814Self-Attention Multilayer Feature Fusion Based on Long ConnectionChu Yuezhong0Wang Jiaqing1Liu Heng2School of Computer Science and TechnologySchool of Computer Science and TechnologySchool of Computer Science and TechnologyFeature fusion is an important part of building high-precision convolutional neural networks. In the field of image classification, though widely used in processing multiscale features of the same layer and short connections in the same receptive field, feature fusion is rarely used in long connection operations across receptive fields. In order to fuse the high- and low-level features of image classification, a feature fusion module SCFF (selective cross-layer feature fusion) for long connections is designed in this work. The SCFF can connect the long-distance feature maps in different receptive fields in a top-down order and apply the self-attention mechanism to fuse them two by two. The final fusion result is used as the input of the classifier. In order to verify the effectiveness of the model, the image classification experiment was done on a number of typical datasets. The experimental results prove that the model can fit the existing convolutional neural network well and effectively improve the classification accuracy of the convolutional network only at the cost of a small amount of calculation.http://dx.doi.org/10.1155/2022/9973814 |
| spellingShingle | Chu Yuezhong Wang Jiaqing Liu Heng Self-Attention Multilayer Feature Fusion Based on Long Connection Advances in Multimedia |
| title | Self-Attention Multilayer Feature Fusion Based on Long Connection |
| title_full | Self-Attention Multilayer Feature Fusion Based on Long Connection |
| title_fullStr | Self-Attention Multilayer Feature Fusion Based on Long Connection |
| title_full_unstemmed | Self-Attention Multilayer Feature Fusion Based on Long Connection |
| title_short | Self-Attention Multilayer Feature Fusion Based on Long Connection |
| title_sort | self attention multilayer feature fusion based on long connection |
| url | http://dx.doi.org/10.1155/2022/9973814 |
| work_keys_str_mv | AT chuyuezhong selfattentionmultilayerfeaturefusionbasedonlongconnection AT wangjiaqing selfattentionmultilayerfeaturefusionbasedonlongconnection AT liuheng selfattentionmultilayerfeaturefusionbasedonlongconnection |