Identification of Tomato Plant Diseases Using CNN- A Comparative Review

Background: Tomatoes are extensively farmed as a vegetable or fruit in many regions of the country, and many forms of tomato pests and diseases are met over the whole life cycle of tomatoes; thus, early identification and diagnosis of these diseases is critical. The application of several deep learning approaches for identification of plant diseases has recently gained popularity, with encouraging results obtained for different diseases. Problem: The advancement of architectures based on Convolutional Neural Network (CNN) has considerably improved diagnostic accuracy of tomato diseases by using images of tomato leaves. The integration of CNN in the agricultural sector ensures an increased produce of tomato crop in a sustainable manner. However, complexity and performance time of CNN technique is a significant concern. Objective: In most deep learning models, all characteristics generated at various layers are given equal weighting. Significant characteristics should be learned and transferred to higher layers of the network for more exact classification. To improve the CNN performance, reuse and sharing of images (large dataset) is a great work required for tomato disease detection to get high accuracy. Methods: In this paper, we have presented a comprehensive review of twenty recent and notable works which employ CNN-based detection of tomato plant diseases. A comparative analysis of these works is given, taking accuracy as the metric. Results: A complete assessment of Bacterial Canker and Speck, Yellow Leaf Curl, Brown Rugose Fruit, Crown and Root Rot, Early Blight, Bunchy Top, Stolbur disease identification studies employing CNNs in this study. Conclusion: It is observed that CNN scan identify diseases with high accuracy when enough training data is provided. We anticipate that our study will be a useful resource for agricultural disease researchers employing technology for early disease diagnosis and management.