A Comprehensive Image Dataset for Accurate Authentication of Meat Species and Hourly Freshness Using Artificial Intelligence in Food Safety

This collection contains 4,500 ultra‑high‑resolution images (4000 × 4000 pixels) of beef and mutton captured in Dhaka’s wet markets over three days in February 2025. Each sample was photographed at five 12‑hour intervals—from immediately post‑slaughter through 48 hours—to document natural spoilage progression. The images are evenly distributed across ten classes (two species × five time points) and come with detailed annotations. To enrich variability without altering tissue morphology, the dataset has been systematically augmented through rotation, noise addition, zooming, and brightness adjustments, expanding it to 18,000 images. The data support dual tasks: authenticating meat species and grading hourly freshness. A custom convolutional neural network trained on this dataset achieved 99.8% accuracy in species recognition and 98% accuracy in freshness grading, underscoring its value for food safety research, supply chain transparency, and consumer protection.