Combining CNN and LSTM for Precise Detection and Classification of Tomato Speck Disease

Abstract

In tomato crops, a fungus called tomato speck disease can result in considerable output losses. For the disease to be managed and controlled effectively, accurate and prompt diagnosis of the condition is essential. In this paper, we present a hybrid CNN and LSTM model for tomato speck disease detection and multi-classification based on 5 distinct severity levels. 10,000 tomato photos from a large dataset were used to train and test the model, which had a binary classification accuracy of 91.18% for determining whether the illness was present or not and an overall multi-classification accuracy of 94.45% for determining the disease severity level. The suggested method outperforms conventional DL approaches in terms of performance, and because to its high degree of accuracy and resilience, it is ideal for use in real-world applications. The results of this study might have a big impact on how tomato speck disease is identified and classified, which could improve the output and quality of tomato crops in the agricultural sector. Future study could involve enhancing it for usage on edge devices and expanding it to additional plant diseases and crops.