Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
157 tokens/sec
GPT-4o
43 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Identification and Recognition of Rice Diseases and Pests Using Convolutional Neural Networks (1812.01043v3)

Published 3 Dec 2018 in cs.CV

Abstract: An accurate and timely detection of diseases and pests in rice plants can help farmers in applying timely treatment on the plants and thereby can reduce the economic losses substantially. Recent developments in deep learning based convolutional neural networks (CNN) have greatly improved the image classification accuracy. Being motivated by the success of CNNs in image classification, deep learning based approaches have been developed in this paper for detecting diseases and pests from rice plant images. The contribution of this paper is two fold: (i) State-of-the-art large scale architectures such as VGG16 and InceptionV3 have been adopted and fine tuned for detecting and recognizing rice diseases and pests. Experimental results show the effectiveness of these models with real datasets. (ii) Since large scale architectures are not suitable for mobile devices, a two-stage small CNN architecture has been proposed, and compared with the state-of-the-art memory efficient CNN architectures such as MobileNet, NasNet Mobile and SqueezeNet. Experimental results show that the proposed architecture can achieve the desired accuracy of 93.3\% with a significantly reduced model size (e.g., 99\% less size compared to that of VGG16).

Citations (321)

Summary

  • The paper demonstrates that fine-tuning pre-trained CNN models, notably VGG16, achieves 97.12% accuracy in recognizing rice diseases and pests.
  • It compares large-scale CNN architectures with a custom lightweight model, enabling effective deployment in mobile applications.
  • The study highlights the use of two-stage training and diverse field data as crucial for robust agricultural diagnostics.

Identification and Recognition of Rice Diseases and Pests Using Convolutional Neural Networks

The paper details a paper on leveraging convolutional neural networks (CNNs) for the detection and classification of rice diseases and pests, a crucial task for mitigating yield losses in rice cultivation. The authors have employed state-of-the-art CNN architectures, notably VGG16 and InceptionV3, for disease identification from rice plant images. Simultaneously, they propose a novel lightweight CNN architecture tailored for mobile applications, ensuring feasibility in rural deployment where computational resources and connectivity are limited.

Methodological Approach

The paper involves two primary phases: experimentation with established large-scale CNN models and the development of a computationally efficient CNN for mobile use. Initially, data was meticulously collected from rice fields to construct a robust dataset comprising eight categories of rice diseases and pests. The dataset spans multiple environmental conditions and captures intra-class variation to ensure comprehensive model training.

In exploring large-scale architectures, the paper tests VGG16 and InceptionV3 using different training strategies: baseline training, transfer learning, and fine-tuning. Notably, fine-tuning, which adapts pre-learned weights from the ImageNet database to the rice dataset, yielded the highest accuracy, reaching 97.12% with VGG16, underscoring the advantages of pre-trained models in domain-specific tasks.

For mobile applications, the complexity and size of traditional CNN models pose significant challenges. The authors introduce a specialized architecture, "Simple CNN," which utilizes a "two-stage training" methodology crafted from the insights gained from fine-tuning. This approach entails an intermediate training phase using classes divided into discernible symptom groups, facilitating feature extraction from limited parameter models. Simple CNN achieved a notable accuracy of 94.33%, efficiently balancing model size against predictive performance.

Results and Implications

The research demonstrates the efficacy of CNNs, particularly fine-tuned architectures, in detecting rice diseases and pests with high accuracy. The Simple CNN model extends these capabilities to resource-constrained environments, making it viable for integration into mobile applications used by farmers in remote areas. Such advancements illustrate the potential for scalable, AI-driven agricultural diagnostics.

The contrast in performance between large and compact CNN architectures is significant as it propels discourse on optimal model design for real-world applications. While large models benefit from abundant parameters and extensive training data, smaller models necessitate innovative training techniques like two-stage training for performance maximization.

Future Directions

The paper opens avenues for future exploration into automated plant disease detection systems integrating various data types, such as environmental conditions and geospatial data, which could enhance prediction accuracy. Moreover, employing segmentation or object detection frameworks may refine disease localization and classification, further integrating heterogeneous backgrounds.

To summarize, this paper offers valuable insights into the deployment of CNNs for agricultural applications, particularly emphasizing practical adaptability for mobile use in resource-limited settings. It lays a groundwork for continued exploration into efficient, precise, and scalable solutions for enhancing agricultural productivity through technological innovation.