EfficientNet-B4-Ranger: A novel method for greenhouse cucumber disease recognition under natural complex environment

Abstract

The intelligent identification and classification of greenhouse plant diseases is an important research object in smart horticulture. In this study, our main task is to find an efficient method to solve the problem of disease similarity caused by two kinds of diseases occurring in the same leaf and the influence of external light. First, we obtain a cucumber leaf disease dataset in a naturally complex greenhouse background, which includes not only powdery mildew, downy mildew, healthy leaves, but also the combination of powdery mildew and downy mildew. Secondly, we use the current state-of-the-art method EfficientNet to construct a classification model for the above four types, Model accuracy is 97%, and prove that EfficientNet-B4 is the most suitable method for this study. Finally, we constructed a two-classification model of cucumber similar diseases by using EfficientNet-B4 improved with the most state-of-the-art optimizer Ranger, obtained unexpected accuracy (96%). The experimental results show that our improved method has significant effect on the classification of similar diseases of cucumber.

Plant phenotyping relevance

キュウリ葉の画像から病害状態を分類する深層学習手法が研究の中心であり、植物の病害表現型を直接推定しているため。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.