Phenomics

Automated Egg Grading through AI and Computer Vision-Driven Morphometric Analysis

---

Abstract

In the poultry industry, accurately measuring egg size is crucial because it impacts breeding, grading, packaging, and marketing. Conventionally, egg size is taken manually, which is time-consuming, inefficient, prone to errors, and can lead to breakage of eggs due to physical handling. Artificial Intelligence offers a faster and more reliable alternative for the same. Therefore, this study was conducted to test the ability of computer vision to estimate egg dimensions automatically. A deep learning-based approach for automated measurement of egg dimensions (length and width) using YOLO11x-OBB (Oriented Bounding Box) object detection was used. A total of 661 eggs were photographed using a commonly available mobile phone camera, with all 661 eggs annotated and included in the analysis (504 chicken eggs, 157 goose eggs). The dataset was split into training (69.9%), validation (15.0%), and test sets (15.1%) to ensure proper model evaluation. Ground truth measurements were obtained using vernier calipers and recorded in a truth set. The YOLO11x-OBB model was trained with data augmentation techniques including rotation, translation, and color space adjustments. Dimension calculation utilized oriented bounding box coordinates and a reference object for scale calibration, with systematic bias correction applied. Overall, for the validation dataset, the system achieved good performance in length measurement with a mean absolute error (MAE) of 0.236cm and a root mean square error (RMSE) of 0.357 cm. Width measurements showed moderate performance with MAE of 0.474 cm, RMSE of 0.674 cm, and R² of 0.171. The correlations observed between predicted and actual dimensions were 0.965 for length and 0.967 for width. Test performance also showed robust generalization with MAE equal to 0.179 cm and width MAE equal to 0.452 cm. Pearson correlation for length equaled 0.982, and for width it was 0.837. Our study suggests that AI-based methods can complement traditional methods. These can enable rapid, contactless, and real-time measurements. Cheaper variants of sophisticated systems can also fit into automated grading lines, hatcheries, and packaging plants. They can reduce labor and improve efficiency. Breeding programs can also benefit from faster data collection.

Keywords: 2026

How to Cite:

Hamadani, A., Boje, P., Moyon, W., Gulzar, M., Khan, A. & Kumar, R., (2026) “Automated Egg Grading through AI and Computer Vision-Driven Morphometric Analysis”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2285334. doi: https://doi.org/10.31274/wcgalp.23648

Rights: 1