The use of digital images could be a faster and cheaper alternative technique to assess BW, HCW, and body composition of beef cattle. The objective of this study was to develop equations to predict body and carcass weight and body fat content of young bulls using digital images obtained through a Microsoft Kinect device. Thirty-five bulls with an initial BW of 383 (±5.38) kg (20 Black Angus, 390 [±7.48] kg initial BW, and 15 Nellore, 377 [±8.66] kg initial BW) were used. The Kinect sensor, installed on the top of a cattle chute, was used to take infrared light-based depth videos, recorded before the slaughter. For each animal, a quality control was made, running and pausing the video at the moment that the animal was standing with its body and head in line. One frame from recorded videos was selected and used to analyze the following body measurements: chest width, thorax width, abdomen width, body length, dorsal height, and dorsal area. From these body measurements, 23 indexes were generated and tested as potential predictors. The BW and HCW were assessed with a digital scale, whereas empty body fat (EBF) was estimated through ground samples of all tissues. To better understand the relationship among the measurements, the correlations between final BW (488 [±10.4] kg), HCW (287 [±12.5] kg), EBF (14 [±0.610] % empty BW) content, body measurements (taken through digital images), and developed indexes were evaluated. The REG procedure was used to develop the regressions, and the important independent variables were identified using the options STEPWISE and Mallow's Cp in the SELECTION statement. Chest width was the trait most related to weights and the correlations between this measurement and BW and HCW were above 0.85. The analysis of linear regressions between observed and predicted values showed that all models pass through the origin and have a slope of unity (null hypothesis [H]: = 0 and = 1; ≥ 0.993). The models to estimate BW and HCW of Angus and Nellore presented between 0.69 and 0.84 ( < 0.001), whereas from equations to estimate the EBF were lower ( = 0.43-0.45; ≤ 0.006). Index I5 [(chest width) × body length], related to the animal volume, was significant in all models created to estimate BW and HCW, and it explained more than 70% of the variation. This study indicates that digital images taken through a Microsoft Kinect system have the potential to be used as a tool to estimate body and carcass weight of beef cattle.

O processo de tipificação determina o mercado para o qual a carcaça será direcionada, considerando sua qualidade. Atualmente, este processo é desenvolvido de maneira subjetiva, e, portanto, a confiabilidade desse pode ser comprometida. Neste sentido, ferramentas que permitam a avaliação objetiva da carcaça bovina in vivo e post mortem são de grande importância. Diante disso, nossos objetivos neste trabalho são (1) avaliar o efeito da classe sexual sobre a morfologia corporal ou de carcaças em bovinos e investigar as relações entre as medidas biométricas e os escores de acabamento atribuídos às carcaças em frigorifico comercial, (2) desenvolver um algoritmo de segmentação utilizando rede neural convolucional, capaz de realizar a segregação de diferentes tecidos em imagens digitais de carcaças, de forma automática, e, (3) desenvolver equações de predição da espessura de gordura subcutânea em bovinos utilizando meta-análise de dados experimentais. Foram utilizados dados de animais de diferentes composições genéticas, classes sexuais, idades e pesos. Os animais foram aleatoriamente selecionados nos currais de espera dos frigoríficos, identificados, submetidos a jejum de sólidos durante 16 horas, pesados e, concomitantemente, foram coletadas imagens do dorso do animal utilizando câmeras RGB-D. Após o abate, a carcaça de cada animal foi dividida longitudinalmente, em duas meias-carcaças, as quais foram pesadas e os escores visuais de gordura avaliados por profissionais treinados, sendo as carcaças resfriadas a 4 oC por 24 horas. As meias-carcaças foram utilizadas para obtenção de imagens RGB-D. As imagens coletadas foram analisadas para mensuração dos parâmetros biométricos. Os dados foram analisados pelos procedimentos gerais de modelos lineares do SAS 9.0 (Statistical Analysis System Institute, Inc.) e, em seguida, as médias das diferentes classes sexuais foram comparadas pelo teste de Tukey (5%). Uma segunda etapa foi realizada utilizando as imagens coletadas como input para uma rede neural convolucional de segmentação (CNN). Na arquitetura da rede, o backbone convolucional utilizado foi a rede ResNeXt-101 combinada com Feature Pyramid Network. Os resultados obtidos indicam que há influência da classe sexual sobre o peso corporal, peso de carcaça e medidas biométricas. Os escores de gordura foram influenciados por medidas biométricas. Houve efeito do peso corporal no escore final e os animais mais pesados obtiveram os maiores escores de gordura. Comportamento semelhante a este foi verificado para a análise das carcaças, com as maiores médias dos parâmetros biométricos observadas para o escore uniforme. Porém, não houve efeito linear entre o escore de gordura e as medidas biométricas, havendo sobreposição das médias para os menores escores atribuídos às carcaças. No segundo estudo, a rede CNN foi capaz de detectar a proporção de tecidos com uma precisão de 66,5, 42,4 e 11,3%, usando sobreposições (IoU) de 25, 50 e 75%, respectivamente. A análise de imagens pode ser usada para obter medidas biométricas em bovinos e suas carcaças e para estimar o escore final de gordura da carcaça. Além disso, esta pesquisa contribui para a proposição de um método de segmentação de carcaças e tecidos o que pode auxiliar no desenvolvimento de sistemas automatizados de avaliação de carcaças. Palavras-chave: Análise de imagem. Bovino. Carcaça. Gordura subcutânea. Predição.

Digital image analysis is a practical, non‐invasive, and relatively low‐cost tool that may assist in the evaluation of body traits in Nile tilapia, being particularly useful for assessing difficult‐to‐measure variables, such as body areas. In this study, we aimed to estimate variance components and genetic parameters for body areas of Nile tilapia obtained by digital images. The data set comprised body weight (BW) records of 1,917 pond‐reared fish at 366 days of age. Of this total, 656 animals were photographed and subjected to image analysis of trunk area (TA), head area (HA), caudal fin area (CFA) and fillet area (FA). Heritabilities and genetic correlations were estimated through multiple‐trait models based on Bayesian inference. Heritability estimates for BW, TA, HA, CFA and FA were 0.25, 0.23, 0.26, 0.21 and 0.25, respectively. Genetic correlations between the traits were high and positive, ranging from 0.70 to 0.98. We highlight the genetic correlation between BW and TA (rG = 0.98) and FA (rG = 0.97). In view of the observed results, it can be concluded that trunk and fillet areas obtained by digital image analysis can lead to indirect genetic gains in weight and other body areas. In addition, the areas studied have potential as a selection criterion and may assist in studies on changes in the body shape in Nile tilapia.

Context Supplementation during the pre- and post-weaning periods is essential to improve the development of beef heifers in tropical pastures. Especially during the dry period, performance of heifers is limited due to low forage yield and poor nutritional value. Aim Evaluate the effect of supplementation during pre- and post-weaning periods on performance, nutritional, and metabolic characteristics in Nellore heifers under grazing. Methods Fifty-four Nellore heifers were randomly divided between the following four supplementation plans, with two replicates: NN, not supplemented in the pre- and post-weaning dry period; NS, not supplemented in the pre-weaning period and supplemented in the post-weaning dry period; SN, supplemented in the pre-weaning period and not supplemented in the post-weaning dry period; and SS, supplemented in the pre- and post-weaning dry periods; with 14, 13, 13 and 14 calves in each treatment respectively. In the post-weaning dry-to-rainy transition period, all heifers were supplemented. Key results Supplementation increased (P < 0.05) the intake of dry matter and crude protein on Day 56 (pre-weaning phase) and Day 168 (post-weaning dry period). Heifers supplemented during the pre-weaning phase had a higher final body weight (fBW) and average daily gain (ADG) on Day 112 (P < 0.05). Heifers NS and SS had higher fBW and ADG on Day 224 (P < 0.05). On Day 280, fBW were higher (P < 0.05) for heifers NS and SS. Insulin-like growth factor 1 was higher for heifers supplemented in the pre-weaning period on Day 112, and higher for NS and SS heifers on Day 224 (P < 0.05). Albumin concentrations were higher (P < 0.05) for heifers NS and SS on Day 280. Supplementation had no effect on either corpus luteum presence or concentration of progesterone (P > 0.05). Conclusions Supplementation during either pre- or post-weaning phases improved multiple performance, nutritional, and metabolic characteristics. The results due to supplementation post-weaning were independent of supplementation pre-weaning. However, supplementation did not result in an improved response to the puberty induction protocol used in this experiment. Implications For replacement heifers, it is important to maintain high weight gains at all stages of growth.