Research Article Open Access

Assessment of Body Condition Scores of Holstein Friesian Crossbred Cows Based on Deep Learning

Abstract

Body Condition Score (BCS) is a measure of body fat or stored energy in the dairy cow. It is an important tool in farm management for achieving better health of cows, reproductive performance and milk yield. Traditionally, BCS is performed visually by veterinary experts, which is time-consuming and involves high cost. Therefore, this study proposed a system based on Convolutional Neural Network (CNN) to automate BCS of cows by image analysis. GNU Image Manipulation Program (GIMP) software was used to remove the background of digitally-captured images of cows, and a MATLAB script was implemented to detect their edges. Finally, the edge-detected images were used as input dataset for the development of deep learning models based on CNN. The image dataset was classified into two groups based on the incremental BCS system of 0.25 (CNN model 1) and 0.5 (CNN model 2). The classification accuracy of the first model for 0.25 and 0.50 error ranges was 63.23% and 85.29%, respectively. In comparison, the second model achieved classification accuracy of 86.02% and 94.85%, for the respective error ranges. Based on the results, the CNN models performed adequately for the middle range of BCS scores wherein the data, most of the cows are present. Therefore, the developed models would perform effectively for commercial dairy farms which do not commonly have cows with poor or high BCS as they would not be very productive.

Shriramulu*, Heartwin A. Pushpadass, Magdaline Eljeeva Emerald Franklin, Manimala Kanagaraj, Jeyakumar Sakthivel, Sivaram Muniandy, Ramesha P. Kerekoppa

To read the full article Download Full Article