Automated computer vision system to predict body weight and average daily gain in beef cattle during growing and finishing phases

Author(s)

Cominotte, A. ^[1] ; Fernandes, A. F. A. ^[1] ; Dorea, J. R. R. ^[1] ; Rosa, G. J. M. ^{[1, 2]} ; Ladeira, M. M. ^[3] ; van Cleef, E. H. C. B. ^[4] ; Pereira, G. L. ^[5] ; Baldassini, W. A. ^[5] ; Machado Neto, O. R. ^{[5, 6]}

Affiliation

^[1] Univ Wisconsin, Dept Anim Sci, Madison, WI 53706 - USA

^[2] Univ Wisconsin, Dept Biostat & Med Informat, Madison, WI 53706 - USA

^[3] Univ Fed Lavras, Anim Sci Dept, BR-3720000 Lavras, MG - Brazil

^[4] Univ Fed Triangulo Mineiro, BR-38280000 Iturama, MG - Brazil

^[5] Sao Paulo State Univ, Sch Vet Med & Anim Sci, BR-18618681 Botucatu, SP - Brazil

^[6] Sao Paulo State Univ, Sch Agr & Veterinarian Sci, BR-14884900 Jaboticabal - Brazil

• Document type: Journal article
• Source: LIVESTOCK SCIENCE; v. 232, FEB 2020.
• Web of Science Citations: 0

Abstract

Frequent measurements of body weight (BW) in livestock systems are very important because they allow assessing growth. However, real-time monitoring of animal growth through traditional weighing scales is stressful for animals, costly and labor-intensive. Thus, the objectives of this study were to: 1) assess the predictive quality of an automated computer vision system used to predict BW and average daily gain (ADG) in beef cattle; and 2) compare different predictive approaches, including Multiple Linear Regression (MLR), Least Absolute Shrinkage and Selection Operator (LASSO), Partial Least Squares (PLS), and Artificial Neutral Networks (ANN). A total of 234 images of Nellore beef cattle were collected during the weaning, stocker and feedlot phases. First, biometric body measurements of each animal, such as body volume, area, length, and others, were performed using three-dimensional images captured with the Kinecto (R) sensor, and their respective BW were acquired using an electronic scale. Next, the biometric measurements were used as explanatory variables in the four predictive approaches (MLR, LASSO, PLS, and ANN). To evaluate prediction quality, a leave-one-out cross-validation was adopted. The ANN was the best prediction approach in terms of Root Mean Square Error of Prediction (RMSEP) and squared predictive correlation (r(2)). The results for Weaning were RMSEP = 8.6 kg and r(2) = 0.91; for Stocker phase, RMSEP = 11.4 kg and r(2) = 0.79; and for Beginning of feedlot, RMSEP = 7.7 kg and r(2) = 0.92. The ANN was also the best method for prediction of ADG, with RMSEP = 0.02 kg/d and r(2) = 0.67 for the period between Weaning and Stocker, RMSEP = 0.02 kg/d and r(2) = 0.85 for the Weaning and Beginning of Feedlot phase, RMSEP = 0.03 kg/d and r(2) = 0.80 for Weaning and Final of Feedlot phase, RMSEP = 0.10 kg/d and r(2) = 0.51 for Stocker and Beginning of feedlot phase, and RMSEP = 0.09 kg/d and r(2) = 0.82 for the Beginning and Final of feedlot phase. Overall, the results indicate that the proposed automated computer vision system can be successfully used to predict BW and ADG in real-time in beef cattle. (AU)

• FAPESP's process: 17/20812-0 - Prediction of body weight and hot carcass weight of Nellore cattle using digital images
• Grantee: Alexandre Cominotte
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• FAPESP's process: 17/02057-0 - Use of visual scores and biometric image in Nellore cattle
• Grantee: Alexandre Cominotte
• Support Opportunities: Scholarships in Brazil - Master