•  
  •  
 

Corresponding Author

Joshua Sokowonci Mommoh

Subject Area

Electronics and Communication Engineering

Article Type

Original Study

Abstract

The development of this platform addresses the challenges associated with traditional tomato disease detection methods as well as the challenges of current machine learning and deep learning models in accurately detecting tomato diseases. Conventional methods of tomato disease diagnosis typically rely on expert evaluation but are time-consuming and susceptible to human mistakes. Incorrect diagnoses of certain diseases could lead to poor control measures, resulting in decreased agricultural productivity. In this research, the PlantVillage dataset, comprising 18,160 tomato leaf images distributed across 10 classes, was used to train two machine learning models, namely, Inception-ResNetV2 and Inception V3. The models were configured, rigorously trained, and evaluated, and their performance was compared. The Inception-ResNetV2 outperformed the Inception V3 by achieving an impressive accuracy of 99.09%, precision of 99.19%, recall of 98.22%, and F1-score of 98.71%. Furthermore, evaluation of models on communication metrics showed that Inception- ResNetV2 had a latency of 47 ms per image and an energy consumption rate of 1.2 W per inference, making it an ideal choice for deployment in resource-constrained environments. Based on the evaluation results, Inception-ResNetV2 was deployed into a web application. Evaluation of the Inception-ResNetV2-powered web application showed that it achieved an average confidence rate of 95%, a data rate of 101.52 KB/s, a response time of 0.161 s, and a throughput of 6.35 images/s.

Keywords

Tomato Disease Detection; Machine Learning; Inception-ResNetV2; Web Application Deployment, ; Agricultural Productivity

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS