Plant Disease Detection through the Implementation of Diversified and Modified Neural Network Algorithms

Abstract

In the era of artificial systems, disease detection is becoming easier. For detecting disease, monitoring the plants 24 hours, visiting the agricultural office, or asking for help from a specialist seem difficult. This situation demands a user-friendly plant disease detection system, which allows people to detect whether the plant is diseased or not in an easier way. If the plant is diseased, a treatment plan will also be notified. In this way, people can easily save time, money, and, most importantly, plants. In this study, the researchers have collected data of vegetables from a field and applied multiple diversified Neural Network Algorithms such as CNN, MCNN, FRCNN, and, along with that, also proposed a new modified neural network architecture (ModCNN), which has produced 97.69% accuracy. The authors have also classified the bean leaf diseases into four categories according to their symptoms, which will help to identify diseases accurately.

Plant phenotyping relevance

植物の葉の症状に基づく病害分類を対象とし、複数のニューラルネットワーク比較と新規ModCNNの提案・評価が研究の中心であるため、植物病害表現型の計算的取得手法として含める。

Code and data availability

The paper describes a custom dataset of 1043 bean and tomato leaf images collected from Sher-E-Bangla Agricultural University and a proposed ModCNN model, but no blocks contain any public dataset deposit, code release, model checkpoint, or availability statement with an authors' URL. The image collection is described,

No evidence-backed public reproduction asset is currently recorded.