Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers
The complex backgrounds of crop disease images and the small contrast between the disease area and the background can easily cause confusion, which seriously affects the robustness and accuracy of apple disease- identification models. To solve the above problems, this paper proposes a Vision Transfo...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-06-01
|
| Series: | Agriculture |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2077-0472/12/6/884 |
| _version_ | 1797490919135510528 |
|---|---|
| author | Xiaopeng Li Shuqin Li |
| author_facet | Xiaopeng Li Shuqin Li |
| author_sort | Xiaopeng Li |
| collection | DOAJ |
| description | The complex backgrounds of crop disease images and the small contrast between the disease area and the background can easily cause confusion, which seriously affects the robustness and accuracy of apple disease- identification models. To solve the above problems, this paper proposes a Vision Transformer-based lightweight apple leaf disease- identification model, ConvViT, to extract effective features of crop disease spots to identify crop diseases. Our ConvViT includes convolutional structures and Transformer structures; the convolutional structure is used to extract the global features of the image, and the Transformer structure is used to obtain the local features of the disease region to help the CNN see better. The patch embedding method is improved to retain more edge information of the image and promote the information exchange between patches in the Transformer. The parameters and FLOPs (Floating Point Operations) of the model are significantly reduced by using depthwise separable convolution and linear-complexity multi-head attention operations. Experimental results on a complex background of a self-built apple leaf disease dataset show that ConvViT achieves comparable identification results (96.85%) with the current performance of the state-of-the-art Swin-Tiny. The parameters and FLOPs are only 32.7% and 21.7% of Swin-Tiny, and significantly ahead of MobilenetV3, Efficientnet-b0, and other models, which indicates that the proposed model is indeed an effective disease-identification model with practical application value. |
| first_indexed | 2024-03-10T00:39:56Z |
| format | Article |
| id | doaj.art-13bb872cc41443948c2a7ccd1fee8a3c |
| institution | Directory Open Access Journal |
| issn | 2077-0472 |
| language | English |
| last_indexed | 2024-03-10T00:39:56Z |
| publishDate | 2022-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Agriculture |
| spelling | doaj.art-13bb872cc41443948c2a7ccd1fee8a3c2023-11-23T15:08:14ZengMDPI AGAgriculture2077-04722022-06-0112688410.3390/agriculture12060884Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on TransformersXiaopeng Li0Shuqin Li1College of Information Engineering, Northwest A&F University, Xianyang 712100, ChinaCollege of Information Engineering, Northwest A&F University, Xianyang 712100, ChinaThe complex backgrounds of crop disease images and the small contrast between the disease area and the background can easily cause confusion, which seriously affects the robustness and accuracy of apple disease- identification models. To solve the above problems, this paper proposes a Vision Transformer-based lightweight apple leaf disease- identification model, ConvViT, to extract effective features of crop disease spots to identify crop diseases. Our ConvViT includes convolutional structures and Transformer structures; the convolutional structure is used to extract the global features of the image, and the Transformer structure is used to obtain the local features of the disease region to help the CNN see better. The patch embedding method is improved to retain more edge information of the image and promote the information exchange between patches in the Transformer. The parameters and FLOPs (Floating Point Operations) of the model are significantly reduced by using depthwise separable convolution and linear-complexity multi-head attention operations. Experimental results on a complex background of a self-built apple leaf disease dataset show that ConvViT achieves comparable identification results (96.85%) with the current performance of the state-of-the-art Swin-Tiny. The parameters and FLOPs are only 32.7% and 21.7% of Swin-Tiny, and significantly ahead of MobilenetV3, Efficientnet-b0, and other models, which indicates that the proposed model is indeed an effective disease-identification model with practical application value.https://www.mdpi.com/2077-0472/12/6/884identification of apple diseasesimage classificationlightweight modelVision Transformerhybrid modelcomplex environments |
| spellingShingle | Xiaopeng Li Shuqin Li Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers Agriculture identification of apple diseases image classification lightweight model Vision Transformer hybrid model complex environments |
| title | Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers |
| title_full | Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers |
| title_fullStr | Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers |
| title_full_unstemmed | Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers |
| title_short | Transformer Help CNN See Better: A Lightweight Hybrid Apple Disease Identification Model Based on Transformers |
| title_sort | transformer help cnn see better a lightweight hybrid apple disease identification model based on transformers |
| topic | identification of apple diseases image classification lightweight model Vision Transformer hybrid model complex environments |
| url | https://www.mdpi.com/2077-0472/12/6/884 |
| work_keys_str_mv | AT xiaopengli transformerhelpcnnseebetteralightweighthybridapplediseaseidentificationmodelbasedontransformers AT shuqinli transformerhelpcnnseebetteralightweighthybridapplediseaseidentificationmodelbasedontransformers |