RealApple-8K: A Real-World Dataset for Apple Freshness Classification

RealApple-8K is a balanced real-world dataset for apple freshness classification containing 8,000 images, including 4,000 fresh and 4,000 rotten apples. The images were captured by the authors using an Apple iPhone 16 in Gorakhpur, Uttar Pradesh, India, between August 2025 and June 2026. The dataset represents diverse real-world visual conditions, including variations in illumination, background clutter, viewpoint, scale, orientation, and partial occlusion. Images were collected under diverse indoor and outdoor conditions without enforcing a standardized imaging setup. The dataset includes more than 100 distinct fresh apple specimens and more than 100 distinct rotten apple specimens. Labels were assigned based on the visible freshness or deterioration condition of the apples and were reviewed by the authors. RealApple-8K is intended to support research on apple freshness classification, real-world image classification, robustness evaluation, and cross-domain benchmarking. The dataset is released to support reproducible research and future evaluation under variable visual conditions.