Scholarlink

In order to safeguard the yield of crops and secure food availability, early plant leaf detection is very important. Over the past few years, deep learning techniques have considerably better image-based disease identification models. Most models are built around CNNs i.e., Convolutional neural networks, that have achieved better accuracy in classifying plant leaf diseases as compared to conventional methods. But CNNs face limitations, including being restricted to cover global features, lack of interpretability, and complexities in generalization across various types of plant species and their conditions. Recently, evolving architectures like Vision Transformers (ViT), BERT-inspired models, and explainable AI (XAI) techniques have come up, which can help tackle the said limitations. This paper provides a detailed review of research articles from 2016 to 2024 focused on plant leaf detection using image-based deep learning. This research studies the evolution of CNN-based models and also explores the shift towards the transformers while evaluating the incorporation of the explainable (XAI) technique in model interpretation. To conclude, it emphasizes the existing research gaps and suggests the future roadmaps for creating effective, interpretable, and scalable detection systems for diseases in plant leaves.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Plant Leaf Disease Detection with Deep Learning Techniques: A Review with a Focus on Transformers and Explainable AI

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Monika Rathore,
  • Harshita Bhati

摘要

In order to safeguard the yield of crops and secure food availability, early plant leaf detection is very important. Over the past few years, deep learning techniques have considerably better image-based disease identification models. Most models are built around CNNs i.e., Convolutional neural networks, that have achieved better accuracy in classifying plant leaf diseases as compared to conventional methods. But CNNs face limitations, including being restricted to cover global features, lack of interpretability, and complexities in generalization across various types of plant species and their conditions. Recently, evolving architectures like Vision Transformers (ViT), BERT-inspired models, and explainable AI (XAI) techniques have come up, which can help tackle the said limitations. This paper provides a detailed review of research articles from 2016 to 2024 focused on plant leaf detection using image-based deep learning. This research studies the evolution of CNN-based models and also explores the shift towards the transformers while evaluating the incorporation of the explainable (XAI) technique in model interpretation. To conclude, it emphasizes the existing research gaps and suggests the future roadmaps for creating effective, interpretable, and scalable detection systems for diseases in plant leaves.