AM-DenseNet:A Novel DenseNet Framework using Attention Mechanisms for COVID-19 CT Image Classification
8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022
; : 474-479, 2022.
Article
in English
| Scopus | ID: covidwho-2281146
ABSTRACT
We present a novel DenseNet framework with attention mechanisms (AM-DenseNet) to extract lung feature of 1 COVID-19 patient. In AM-DenseNet, a lightweight Efficient Channel Attention (ECA) structure is added at the end of each dense connection to introduce an attention mechanism to discovery local lesion domain. We compare our AM-DenseNet to VGG-16, ResNet-50 and DenseNet-121 on CT image dataset of COVID-19 patients. According to the experimental results, we conclude that the classification performance of AM-DensNet framework can be significantly enhanced under the effect of attention mechanism. The AM-DensNet shows better classification performance than the compared models. © 2022 IEEE.
Full text:
Available
Collection:
Databases of international organizations
Database:
Scopus
Language:
English
Journal:
8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022
Year:
2022
Document Type:
Article
Similar
MEDLINE
...
LILACS
LIS