Abstract: Tomato is one of the most extensively grown vegetables in any country, and their diseases can significantly affect yield and quality. Accurate and early detection of tomato diseases is crucial for reducing losses and improving crop management. Current Deep Learning and CNN research have resulted in the availability of multiple CNN designs, making automated plant disease identification viable rather than traditional visual inspection-based disease detection. When using Deep Learning Methods, the dataset serves one of the most crucial roles in disease prediction. PlantVillage is the most widely used publicly available dataset for Tomato Disease detection, but it was created in a lab/controlled environment, and models trained on it do not perform well on real-world images. Some natural or real-world datasets are available, but they are private and not publicly available. Also, when attempting to predict tomato diseases on the field in the Jodhpur and Jaipur districts of Rajasthan, India, we found that the majority of diseases are Leaf Miner, spotted wilt virus, and Nutrition deficiency diseases, but there are no public datasets containing such categories. To overcome these challenges, we propose the creation of a new dataset called "Tomato-Village" with three variants: a) Multiclass tomato disease classification, b) Multilabel tomato disease classification and c) Object detection based tomato disease detection. As per our best knowledge, “Tomato-Village” will be the first such dataset to be available publicly. Further, we have applied the various CNN architectures/models on this dataset, and baseline results are drawn.
Gehlot, M., Saxena, R.K. & Gandhi, G.C. “Tomato-Village”: a dataset for end-to-end tomato disease detection in a real-world environment. Multimedia Systems (2023). DOI : https://doi.org/10.1007/s00530-023-01158-y
Article Link : https://link.springer.com/article/10.1007/s00530-023-01158-y