CowAnalyzer: visual features based smart breed identification application

In Pakistan, a huge portion of the agricultural Gross Domestic Product (GDP) depends on the trade of high-quality breeds which is beneficial for the livestock sector. However, more than 80% of livestock breeds in Pakistan are non-descriptive, which burdens the economy as we are unaware of their potential. Livestock breed identification is thus important for tractability, animal trading, and enhancing GDP growth. Therefore in this work, we proposed a visual features-based smart breed identification application called CowAnalyzer. The CowAnalyzer performs fast, easy-to-use, cost-effective, and reliable cow breed variates classification. It performs visual feature engineering of 10 Pakistan’s cattle breeds, by utilizing morphological, color, pattern, structure, shape, and external organ-specific features. The feature engineering is then utilized by the AI model for the identification of breeds. The CowAnalyzer achieved an overall accuracy of 79% with a moderate success rate for all breeds. The CowAnalyzer is deployed on the PakSupercomputing cluster which provides an easy-to-access and valuable platform for livestock industries including cattle farms and dairy farms.