GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition
Lightweight convolutional neural network (CNN) models have proven effective in recognizing common pest species, yet challenges remain in enhancing their nonlinear learning capacity and reducing overfitting. This study introduces a grouped dropout strategy and modifies the CNN architecture to improve...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-10-01
|
| Series: | Agriculture |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2077-0472/14/11/1915 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846154878552899584 |
|---|---|
| author | Dongcheng Li Yongqi Xu Zheming Yuan Zhijun Dai |
| author_facet | Dongcheng Li Yongqi Xu Zheming Yuan Zhijun Dai |
| author_sort | Dongcheng Li |
| collection | DOAJ |
| description | Lightweight convolutional neural network (CNN) models have proven effective in recognizing common pest species, yet challenges remain in enhancing their nonlinear learning capacity and reducing overfitting. This study introduces a grouped dropout strategy and modifies the CNN architecture to improve the accuracy of multi-class insect recognition. Specifically, we optimized the base model by selecting appropriate optimizers, fine-tuning the dropout probability, and adjusting the learning rate decay strategy. Additionally, we replaced ReLU with PReLU and added BatchNorm layers after each Inception layer, enhancing the model’s nonlinear expression and training stability. Leveraging the Inception module’s branching structure and the adaptive grouping properties of the WeDIV clustering algorithm, we developed two grouped dropout models, the iGDnet-IP and GDnet-IP. Experimental results on a dataset containing 20 insect species (15 pests and five beneficial insects) demonstrated an increase in cross-validation accuracy from 84.68% to 92.12%, with notable improvements in the recognition rates for difficult-to-classify species, such as <i>Parnara guttatus</i> Bremer and Grey (PGBG) and <i>Papilio xuthus</i> Linnaeus (PXLL), increasing from 38% and 47% to 62% and 93%, respectively. Furthermore, these models showed significant accuracy advantages over standard dropout methods on test sets, with faster training times compared to four conventional CNN models, highlighting their suitability for mobile applications. Theoretical analyses of model gradients and Fisher information provide further insight into the grouped dropout strategy’s role in improving CNN interpretability for insect recognition tasks. |
| format | Article |
| id | doaj-art-e8bdea3b0c83423482ab426935dc4dfc |
| institution | Kabale University |
| issn | 2077-0472 |
| language | English |
| publishDate | 2024-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Agriculture |
| spelling | doaj-art-e8bdea3b0c83423482ab426935dc4dfc2024-11-26T17:43:19ZengMDPI AGAgriculture2077-04722024-10-011411191510.3390/agriculture14111915GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest RecognitionDongcheng Li0Yongqi Xu1Zheming Yuan2Zhijun Dai3Hunan Engineering & Technology Research Center for Agricultural Big Data Analysis & Decision-Making, Hunan Agricultural University, Changsha 410128, ChinaHunan Engineering & Technology Research Center for Agricultural Big Data Analysis & Decision-Making, Hunan Agricultural University, Changsha 410128, ChinaHunan Engineering & Technology Research Center for Agricultural Big Data Analysis & Decision-Making, Hunan Agricultural University, Changsha 410128, ChinaHunan Engineering & Technology Research Center for Agricultural Big Data Analysis & Decision-Making, Hunan Agricultural University, Changsha 410128, ChinaLightweight convolutional neural network (CNN) models have proven effective in recognizing common pest species, yet challenges remain in enhancing their nonlinear learning capacity and reducing overfitting. This study introduces a grouped dropout strategy and modifies the CNN architecture to improve the accuracy of multi-class insect recognition. Specifically, we optimized the base model by selecting appropriate optimizers, fine-tuning the dropout probability, and adjusting the learning rate decay strategy. Additionally, we replaced ReLU with PReLU and added BatchNorm layers after each Inception layer, enhancing the model’s nonlinear expression and training stability. Leveraging the Inception module’s branching structure and the adaptive grouping properties of the WeDIV clustering algorithm, we developed two grouped dropout models, the iGDnet-IP and GDnet-IP. Experimental results on a dataset containing 20 insect species (15 pests and five beneficial insects) demonstrated an increase in cross-validation accuracy from 84.68% to 92.12%, with notable improvements in the recognition rates for difficult-to-classify species, such as <i>Parnara guttatus</i> Bremer and Grey (PGBG) and <i>Papilio xuthus</i> Linnaeus (PXLL), increasing from 38% and 47% to 62% and 93%, respectively. Furthermore, these models showed significant accuracy advantages over standard dropout methods on test sets, with faster training times compared to four conventional CNN models, highlighting their suitability for mobile applications. Theoretical analyses of model gradients and Fisher information provide further insight into the grouped dropout strategy’s role in improving CNN interpretability for insect recognition tasks.https://www.mdpi.com/2077-0472/14/11/1915common pestslightweight CNNinsect image recognitiongrouped dropoutactivation functionbatch normalization |
| spellingShingle | Dongcheng Li Yongqi Xu Zheming Yuan Zhijun Dai GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition Agriculture common pests lightweight CNN insect image recognition grouped dropout activation function batch normalization |
| title | GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition |
| title_full | GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition |
| title_fullStr | GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition |
| title_full_unstemmed | GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition |
| title_short | GDnet-IP: Grouped Dropout-Based Convolutional Neural Network for Insect Pest Recognition |
| title_sort | gdnet ip grouped dropout based convolutional neural network for insect pest recognition |
| topic | common pests lightweight CNN insect image recognition grouped dropout activation function batch normalization |
| url | https://www.mdpi.com/2077-0472/14/11/1915 |
| work_keys_str_mv | AT dongchengli gdnetipgroupeddropoutbasedconvolutionalneuralnetworkforinsectpestrecognition AT yongqixu gdnetipgroupeddropoutbasedconvolutionalneuralnetworkforinsectpestrecognition AT zhemingyuan gdnetipgroupeddropoutbasedconvolutionalneuralnetworkforinsectpestrecognition AT zhijundai gdnetipgroupeddropoutbasedconvolutionalneuralnetworkforinsectpestrecognition |