Issue |
ITM Web Conf.
Volume 70, 2025
2024 2nd International Conference on Data Science, Advanced Algorithm and Intelligent Computing (DAI 2024)
|
|
---|---|---|
Article Number | 03025 | |
Number of page(s) | 7 | |
Section | Image Processing and Computer Vision | |
DOI | https://doi.org/10.1051/itmconf/20257003025 | |
Published online | 23 January 2025 |
Dual Convolution Neural Networks of Ensemble Learning with Attention Mechanism for Rice Classification
Computer Sciences, Chengdu University of Technology, 610059 Chengdu, China
Corresponding author: cheng.linfeng1@student.zy.cdut.edu.cn
Machine vision has been widely applied across fields. Image classification is one of the most classic fields. The aim of this project is to develop a dual convolutional neural network for ensemble learning based on the initial model and the res network model, and apply the ensemble model to the rice classification problem. The ensemble model in this article combines two deep models, InceptionNet and ResNet, and incorporates self-attention block method to construct an attention mechanism that uses multi head attention layers to capture relationships in the input. Attention output is added back to input. At the same time, 10 evaluation indicators were introduced as the results of testing and evaluation. In the result analysis, it can be concluded that the ensemble model has demonstrated excellent training efficiency in these indicators, and the learning rate hyperparameter has been replaced to improve the stability of the model. At the same time, for a more comprehensive comparison, the ensemble model studied in this article was also compared and analyzed reasonably with three pre trained models: VGG-16, ResNet50, and MobileNet. In the future, it is necessary to continuously optimize the structure of integrated models and adjust their hyperparameters to achieve better stability.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.