Multiclass Image Dataset of Tomato Pathologies: Integrating Leaf and Fruit Samples for Interpretable Deep Learning

Description: This dataset provides a high-resolution, curated collection of 4,120 annotated images of tomato (Solanum lycopersicum) leaves and fruits. Specifically designed to support the development of Explainable AI (XAI) models in precision agriculture, the dataset captures a holistic view of plant pathology. By offering balanced data across both foliage and fruit, it enables researchers and AI practitioners to build more transparent and comprehensive crop health monitoring systems that account for the entire plant's physiological state.

Dataset Content: The collection encompasses 4,120 images organized into four distinct, balanced classes. This structure ensures robust training for image-based classification and localization tasks, focusing on the visual manifestations of fungal, bacterial, and viral infections compared against healthy baselines.

Healthy Leaf (1,013 images): Visuals of foliage showing no signs of infection, stress, or nutrient deficiency.

Infected Leaf (1,050 images): Foliage displaying diverse pathological symptoms, including Early Blight, Late Blight, and Leaf Mold.

Healthy Fruit (1,025 images): Ripe and unripe tomatoes characterized by clear skin and the absence of lesions.

Infected Fruit (1,065 images): Tomatoes exhibiting visible disease symptoms, surface lesions, or active rot.

Purpose: The primary objective of this dataset is to bridge the gap between black-box deep learning and interpretable agricultural diagnostics. By providing high-resolution data for both leaves and fruit, it facilitates the creation of models that can justify their diagnostic decisions. This supports advancements in precision farming, automated disease management, and crop yield optimization, ultimately contributing to more resilient food production systems.