Image Dataset for Tomato Quality Grading: Healthy vs. Defective Classes for Deep Learning

This dataset contains a collection of tomato images captured under uncontrolled, real-world conditions. Unlike laboratory-grade datasets, these images were taken using various mobile phone cameras in everyday environments, featuring tomatoes purchased from local supermarkets.

The core objective of this dataset is to provide a "real-world" benchmark for image classification, forcing models to handle variability in lighting, backgrounds, and camera quality.

Dataset Structure:

The data is organized into a standard directory structure for Deep Learning:

train/: Images for model training.

validation/: Images for model evaluation and hyperparameter tuning.

Classes and Labeling:

The dataset is split into two categories based on visual quality:

tomates_ok: Tomatoes in good condition. Filenames follow the pattern *_1.jpg.

tomates_no_ok: Tomatoes with visual defects, damage, or in poor condition. Filenames follow the pattern *_0.jpg.

Technical Context:

Source: Real-world retail environments (supermarkets).

Capture Device: Various mobile smartphones (non-standardized).

Environment: Uncontrolled lighting, diverse backgrounds, and varying angles.

Format: JPG.

Task: Binary Image Classification for automated quality inspection.