Comparative analysis of image augmentation methods for improving the quality of computer vision models in animal husbandry
Irina I. Mikhailenko, Cand. Sc. (Biology), researcher, Laboratory of Intellectual Agricultural Technologies. Belgorod Federal Research Center of the Russian Academy of Sciences
Abstract
The research is aimed at developing robust computer vision models for precision animal husbandry tasks (health monitoring, animal identification, and body condition scoring), which are commonly hampered by a scarcity of labeled data. Collecting and annotating images in real-life farm conditions is an expensive and labor-intensive process. Data augmentation methods are a key solution for expanding training sets and improving the generalization capability of models. This paper examines several methods for preprocessing cattle image data and explores augmentation methods that use the pre-existing data to generate new data and enrich training data without human intervention, thereby enhancing performance. The experiments have addressed various aspects of data scarcity, yielding competitive results that provide a baseline for future research.
Keywords: augmentation strategies; image transformation; machine learning; neural networks; limited data.
For citation: Mikhailenko I. I. Comparative analysis of image augmentation methods for improving the quality of computer vision models in animal husbandry. Digital Models and Solutions. 2026. Vol. 5, no. 3, pp. 5–14. DOI: 10.29141/2949-477X-2026-5-3-1. EDN: FCGAES.
