Swin-Transformer Based Classification for Rice Diseases Recognition

Zhang, Zhenghua; Gong, Zhangjie; Hong, Qingyang; Jiang, Ling

doi:10.1109/cisai54367.2021.00036

Cited by 16 publications

(6 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, this hierarchical converter is capable of modeling images of various sizes and has linear computation complexity. As a result of these features, Swin converter is highly competitive in handling a wide variety of visual tasks ( Zhang et al, 2021 ; Jiang et al, 2022 ).…”

Section: Methodsmentioning

confidence: 99%

FunSwin: A deep learning method to analysis diabetic retinopathy grade and macular edema risk based on fundus images

et al. 2022

View full text Add to dashboard Cite

Diabetic retinopathy (DR) and age-related macular degeneration (AMD) are forms of degenerative retinal disorders that may result in vision impairment or even permanent blindness. Early detection of these conditions is essential to maintaining a patient’s quality of life. The fundus photography technique is non-invasive, safe, and rapid way of assessing the function of the retina. It is widely used as a diagnostic tool for patients who suffer from fundus-related diseases. Using fundus images to analyze these two diseases is a challenging exercise, since there are rarely obvious features in the images during the incipient stages of the disease. In order to deal with these issues, we have proposed a deep learning method called FunSwin. The Swin Transformer constitutes the main framework for this method. Additionally, due to the characteristics of medical images, such as their small number and relatively fixed structure, transfer learning strategy that are able to increase the low-level characteristics of the model as well as data enhancement strategy to balance the data are integrated. Experiments have demonstrated that the proposed method outperforms other state-of-the-art approaches in both binary and multiclass classification tasks on the benchmark dataset.

show abstract

Section: Methodsmentioning

confidence: 99%

FunSwin: A deep learning method to analysis diabetic retinopathy grade and macular edema risk based on fundus images

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Li et al [51] introduced a hybrid network model that combines CNN and Transformer for precise image localization. Zhang et al [52] employed a novel rice disease identification method based on the Swin Transformer [53], effectively enhancing the accuracy of rice disease detection.…”

Section: Research On Machine Learning For Architectural Classificationmentioning

confidence: 99%

Classification and Model Explanation of Traditional Dwellings Based on Improved Swin Transformer

Miao,

Zhang,

Piao

et al. 2024

Buildings

View full text Add to dashboard Cite

The extraction of features and classification of traditional dwellings plays significant roles in preserving and ensuring the sustainable development of these structures. Currently, challenges persist in subjective classification and the accuracy of feature extraction. This study focuses on traditional dwellings in Gansu Province, China, employing a novel model named Improved Swin Transformer. This model, based on the Swin Transformer and parallel grouped Convolutional Neural Networks (CNN) branches, aims to enhance the accuracy of feature extraction and classification precision. Furthermore, to validate the accuracy of feature extraction during the prediction process and foster trust in AI systems, explainability research was conducted using Grad-CAM-generated heatmaps. Initially, the Gansu Province Traditional Dwelling Dataset (GTDD) is established. On the constructed GTDD dataset, the Improved Swin Transformer attains an accuracy of 90.03% and an F1 score of 87.44%. Comparative analysis with ResNet-50, ResNeXt-50, and Swin Transformer highlights the outstanding performance of the improved model. The confusion matrix of the Improved Swin Transformer model reveals the classification results across different regions, indicating that the primary influencing factors are attributed to terrain, climate, and cultural aspects. Finally, using Grad-CAM-generated heatmaps for explaining classifications, it is observed that the Improved Swin Transformer model exhibits more accurate localization and focuses on features compared to the other three models. The model demonstrates exceptional feature extraction ability with minimal influence from the surrounding environment. Simultaneously, through the heatmaps generated by the Improved Swin Transformer for traditional residential areas in five regions of Gansu, it is evident that the model accurately extracts architectural features such as roofs, facades, materials, windows, etc. This validates the consistency of features extracted by the Improved Swin Transformer with traditional methods and enhances trust in the model and decision-making. In summary, the Improved Swin Transformer demonstrates outstanding feature extraction ability and accurate classification, providing valuable insights for the protection and style control of traditional residential areas.

show abstract

“…Through experimental comparison with the CNNs, the identification accuracy of vision transformer-based approach improved by nearly 2%. Zhang et al [12] developed a rice disease identification approach using Swin-transformer and improved the accuracy by 4.1% compared to that achieved using conventional machine learning models. Together, these works greatly advanced our understanding of plant disease identification using CNNs or vision transformers.…”

Section: Obviously Vision Transformer Architectures Have Appeared As ...mentioning

confidence: 99%

Are Transformers more Suitable for Plant Disease Identification than Convolutional Neural Networks?

Zhou,

Zhang,

Zhong

2024

Preprint

View full text Add to dashboard Cite

Recent studies suggest that the transformer-based architectures have made unprecedented achievements in various computer vision tasks, such as image classification, object detection, semantic segmentation, etc. However, as the de facto approach, convolutional neural networks (CNNs) have reigned for a decade in plant disease identification tasks. We cannot help but propose a scenario: are transformers more suitable for plant disease identification than CNNs? Conceivably, this work aims to further investigate and evaluate whether it is feasible to trivially switch to transformers in plant disease identification tasks. To address this issue, a series of experiments were curated into three training strategies and implemented on two representative plant disease image datasets. Experimental results indicated that the vanilla transformer methods performed unsatisfactorily when trained from scratch, and the two mainstream transfer learning-based approaches achieved better prediction performance but with little margin. It is encouraging that the self-supervision-based transformer methods outperform their CNN-based counterparts, which provides new insights for agricultural image processing. Importantly, this work revealed a model selection solution based on the experimental results and target datasets, and gained a better interpretation by visualizing the feature maps.

show abstract

Swin-Transformer Based Classification for Rice Diseases Recognition

Cited by 16 publications

References 5 publications

FunSwin: A deep learning method to analysis diabetic retinopathy grade and macular edema risk based on fundus images

FunSwin: A deep learning method to analysis diabetic retinopathy grade and macular edema risk based on fundus images

Classification and Model Explanation of Traditional Dwellings Based on Improved Swin Transformer

Are Transformers more Suitable for Plant Disease Identification than Convolutional Neural Networks?

Contact Info

Product

Resources

About