A Machine-Vision Framework for Automated Egg Variety Recognition, Freshness Assessment, and Nutritional Estimation Using Machine Learning Techniques
Description
This dataset contains image data developed for automated egg species recognition, storage-duration-based visual freshness assessment, and mass-based nutritional estimation. It includes high-resolution RGB images of five egg categories: Bird Koel, Country Chicken, Red Layer Chicken, White Layer Chicken, and Duck. The images were collected from local markets in Dhaka, Bangladesh, under practical lighting conditions to reflect real-world visual variations in shell color, texture, speckling, gloss, shape, and surface condition. For egg species and freshness recognition, the dataset contains 9,100 original captured egg images organized into five species classes and fifteen species–freshness visual classes. The freshness-related classes were formed from three storage-duration-based stages: Fresh/Day 0, Half-Fresh/Day 25, and Rotten/Day 45. The images were arranged into separate training, validation, and test subsets for model development and final evaluation. No data augmentation was applied to the final classification dataset. The dataset also includes 2,250 egg-and-coin images, where each egg is photographed beside a Bangladeshi 5-taka coin. These images support reference-based egg size, mass, and nutritional estimation using image segmentation and predefined species-level nutrient coefficients. This dataset supports research in computer vision, deep learning, food quality assessment, egg species recognition, visual freshness classification, mass-based nutritional estimation, and non-invasive agricultural product analysis.