Arnes Sembiring, Yuwaldi Away, Fitri Arnia, Rusdha Muharar
Early detection of plant diseases requires a monitoring system that can classify plants accurately and run automatically in the agricultural field. Convolutional Neural Network offers better classification accuracy than various other classification methods but its implementation in agricultural monitoring system faces various challenges. The main challenge is the large amount of resources required to run various existing Convolutional Neural Network architectures. This study proposed a concise Convolutional Neural Network architecture and investigated the performance of its configuration in classifying tomato diseases based on leaf images. The proposed architectural model has a much smaller number of parameters compared to the 5 comparative architectural models in this study. The proposed architecture has 109.8k parameters while Squeezenet and Shufflenet as the most concise comparison CNN architectures in this study have 1.24M and 1.4M parameters, respectively. All comparison models and proposed architecture were trained and tested on the same dataset, namely the image of tomato leaves from PlantVillage Dataset. The results revealed that SqueezeNet was the fastest comparison CNN with 98.70% accuracy while Shufflenet achieved 98.89% accuracy. This accuracy is slightly higher than the proposed architecture, which is 98.64% but the proposed architecture is 7.9 times faster than SqueezeNet and 8.7 times than Shufflenet in classifying each image. Classification speed, small number of parameters and competitive classification accuracy indicate that the proposed architecture is feasible to be applied in agricultural monitoring system. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
Doctoral Program, School of Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia; Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock