ResViT-Rice: A Deep Learning Model Combining Residual Module and Transformer Encoder for Accurate Detection of Rice Diseases

Abstract

Rice is a staple food for over half of the global population, but it faces significant yield losses: up to 52% due to leaf blast disease and brown spot diseases, respectively. This study aimed at proposing a hybrid architecture, namely ResViT-Rice, by taking advantage of both CNN and transformer for accurate detection of leaf blast and brown spot diseases. We employed ResNet as the backbone network to establish a detection model and introduced the encoder component from the transformer architecture. The convolutional block attention module was also integrated to ResViT-Rice to further enhance the feature-extraction ability. We processed 1648 training and 104 testing images for two diseases and the healthy class. To verify the effectiveness of the proposed ResViT-Rice, we conducted comparative evaluation with popular deep learning models. The experimental result suggested that ResViT-Rice achieved promising results in the rice disease-detection task, with the highest accuracy reaching 0.9904. The corresponding precision, recall, and F1-score were all over 0.96, with an AUC of up to 0.9987, and the corresponding loss rate was 0.0042. In conclusion, the proposed ResViT-Rice can better extract features of different rice diseases, thereby providing a more accurate and robust classification output.

Advancements in rice disease detection through convolutional neural networks: A comprehensive review

“…Examining specific papers reveals the strategic use of established datasets like PlantVillage and Kaggle Rice Disease, as seen in studies by Refs. [ 8 , 57 ], respectively, underlining the importance of leveraging well-established resources. The classification diversity, spanning from 2 to 10 classes, underscores the nuanced and intricate nature of the various rice diseases studied.…”

“…Each entry showcases the thoughtful considerations and techniques applied to enhance the quality and relevance of the datasets before feeding them into the models. Strategies range from basic adjustments, such as size cutting and angle changes [ 88 ] and contrast, brightness, and color adjustments [ 8 ], to more advanced techniques like principal component analysis (PCA) [ 89 ] and dual-tree complex wavelet transform (DTCWT) [ 90 ]. Notable is the variety of approaches employed for image enhancement, including the use of Generative Adversarial Networks (GANs) [ 91 ], Progressive training, PWGAN-GP method, TIDA method, and Test set imbalance adjustment [ 92 ], and the application of Hybrid Gaussian-Weiner (HGW) filters for noise removal [ 93 ].…”

“…The rice industry's impact on global trade and economies, both on small and large scales, underscores its pivotal role in international commerce. Beyond its role as a major food source, rice's cultural significance, economic impact, and contributions to global food security emphasize its versatility, nutritional value, and adaptability in sustaining human life and well-being [ [5] , [6] , [7] , [8] ].…”