Web-Based AI Plant Disease Detection and Treatment Recommendation System using Deep Learning

Abstract

Agriculture remains a fundamental pillar of food security and economic sustainability, particularly in developing countries where a large proportion of the population depends on farming for their livelihood. One of the major challenges faced by the agricultural sector is the occurrence of plant diseases, which significantly reduce crop yield, degrade product quality, and lead to substantial economic losses. Plant diseases caused by fungi, bacteria, viruses, and pests often spread rapidly, and delayed identification can result in large-scale crop damage. Therefore, early and accurate detection of plant diseases is crucial for effective disease management and sustainable agricultural practices. Traditional methods of plant disease identification primarily rely on manual inspection by agricultural experts or laboratory-based diagnostic techniques. Although these methods can provide reliable results, they are time-consuming, labor-intensive, and often inaccessible to small-scale farmers, especially in rural and remote regions. In many cases, farmers lack immediate access to expert guidance, leading to improper disease diagnosis and the excessive or incorrect use of pesticides. Such practices not only reduce crop productivity but also pose serious environmental and health risks. These limitations highlight the need for automated, accessible, and cost-effective plant disease diagnosis solutions. Recent advancements in artificial intelligence (AI), particularly in deep learning and computer vision, have enabled significant progress in automated image-based plant disease detection. Convolutional Neural Networks (CNNs) have demonstrated strong capability in learning discriminative visual features from plant leaf images and achieving high classification accuracy across multiple crop species and disease categories. Transfer learning using pretrained models has further improved performance while reducing training time and computational requirements. However, many existing deep learning-based systems focus mainly on disease classification accuracy and often overlook practical deployment challenges, computational efficiency, and decision-support functionalities required for real-world agricultural applications. Moreover, most current approaches provide only disease labels as output, without offering actionable treatment recommendations or assessing the reliability of predictions. In real-world scenarios, farmers require not only disease identification but also guidance on appropriate organic and chemical control measures to take timely action. The absence of confidence estimation and uncertainty handling in many automated systems can lead to misleading predictions, which may result in inappropriate treatment decisions and further crop damage. Additionally, heavy deep learning architectures often limit the feasibility of deploying such systems in web-based or resource-constrained environments. To address these challenges, this work proposes a web-based intelligent plant disease detection and treatment recommendation system that integrates deep learning and machine learning techniques. A lightweight pretrained CNN model is employed as a feature extractor to capture relevant visual characteristics from plant leaf images, while a machine learning classifier is used for efficient and accurate disease classification. The proposed system further incorporates confidence-based disease severity assessment and provides organic and chemical treatment recommendations through a structured knowledge base. By offering real-time analysis through a user-friendly web interface, the system aims to support farmers in making informed decisions, reduce dependency on expert consultation, and promote timely and sustainable disease management practices.

Plant phenotyping relevance

葉画像から植物病害を検出し、病害重症度を推定する画像・深層学習ワークフローが研究の中心であり、植物状態の取得・推定手法として実質的です。

Code and data availability

The article describes a MobileNetV2-based plant disease detection web system but contains no public dataset, image, code, model, or supplement availability statements with author URLs. PlantVillage is mentioned only as a cited benchmark dataset, not a paper-specific deposit.

No evidence-backed public reproduction asset is currently recorded.