Supervised learning techniques for dairy cattle body weight prediction from 3D digital images

How much can performance measures explain of the between-cow variation in enteric methane?

Enteric CH4 produced from dairy cows contributes to the emission of greenhouse gases from anthropogenic sources. Recent studies have shown that the selection of lower CH4-emitting cows is possible, but doing so would be simpler if performance measures already recorded on farm could be used, instead of measuring gas emissions from individual cows. These performance measures could be used for selection of low emitting cows. The aim of this analysis was to quantify how much of the between-cow variation in CH4 production can be explained by variation in performance measures. A dataset with 3 experiments and a total of 149 lactating dairy cows with repeated measures was used to estimate the between-cow variation (the variation between cow estimates) for performance and gas measures from GreenFeed (C-Lock, Rapid City, SD). The cow estimates were obtained with a linear mixed model with the diet within period effect as a fixed effect and the cow within experiment as a random effect. The cow estimates for CH4 production were first regressed on the performance and gas measures individually, and then performance and CO2 production measures were grouped in 3 subsets for principal component analysis and principal component regression. The variables that explained most of the between-cow variation in CH4 production were DMI (R2 = 0.44), among the performance measures, and CO2 production (R2 = 0.61), among gas measures. Grouping the measures increased the R2 to 0.53 when only performance measures were used, and to 0.66 when CO2 production was added to the significant performance measures. We found the marginal improvement to be insufficient to justify the use of grouped measures rather than an individual measure because the latter simplifies the model and avoids over-fitting. Investigation of other measures that can be explored to increase explanatory power of between-cow variation in CH4 production is briefly discussed. Finally, the use of residual CH4 as a measure for CH4 efficiency could be considered by using either DMI or CO2 production as the sole predicting variables.

Feed intake in housed dairy cows: validation of a three-dimensional camera-based feed intake measurement system

Measuring feed intake accurately is crucial to determine feed efficiency and for genetic selection. A system using three-dimensional (3D) cameras and deep learning algorithms can measure the volume of feed intake in dairy cows, but for now, the system has not been validated for feed intake expressed as weight of feed. The aim of this study was to validate the weight of feed intake predicted from the 3D cameras with the actual measured weight. It was hypothesised that diet-specific coefficients are necessary for predicting changes in weight, that the relationship between weight and volume is curvilinear throughout the day, and that manually pushing the feed affects this relationship. Twenty-four lactating Danish Holstein cows were used in a cross-over design with four dietary treatments, 2 × 2 factorial arranged with either grass-clover silage or maize silage as silage factor, and barley or dried beet pulp as concentrate factor. Cows were adapted to the diets for 11 d, and for 3 d to tie-stall housing before camera measurements. Six cameras were used for recording, each mounted over an individual feeding platform equipped with a weight scale. When building the predictive models, four cameras were used for training, and the remaining two for testing the prediction of the models. The most accurate predictions were found for the average feed intake over a period when using the starting density of the feed pile, which resulted in the lowest errors, 6% when expressed as RMSE and 5% expressed as mean absolute error. A model including curvilinear effects of feed volume and the impact of manual feed pushing was used on a dataset including daily time points. When cross-validating, the inclusion of a curvilinear effect and a feed push effect did not improve the accuracy of the model for neither the feed pile nor the feed removed by the cow between consecutive time points. In conclusion, measuring daily feed intake from this 3D camera system in the present experimental setup could be accomplished with an acceptable error (below 8%), but the system should be improved for individual meal intake measurements if these measures were to be implemented.

Body weight prediction of Belgian Blue crossbred using random forest

Objective

The aim of this study was to predict the body weight (BW) of a Belgian Blue X Friesian Holstein (BB X FH) crossbred in Indonesia based on morphometrics using random forest.

Materials and Methods

A total of 26 BB X FH crossbreds were observed for BW, chest weight (CW), body length (BL), hip height (HH), wither height (WH), and chest girth (CG) from 0, 30, 60, 90, 120, 150, 180, 210, 240, 270, and 300 days of age. Stepwise regression and random forest were performed using R 3.6.1.

Results

The random forest results show that CG is an important variable in estimating BW, with an important variable value of 24.49%. Likewise, the results obtained by stepwise regression show that CG can be an indicator of selection for the BB X FH crossbred. The R squared value obtained from the regression is 0.83, while the R squared value obtained from the random forest (0.86) is greater than the regression.

Conclusion

In conclusion, random forest produces a better model than stepwise regression. However, a good simple equation to use to estimate BW is CG.

Repeatabilities of individual measures of feed intake and body weight on in-house commercial dairy cattle using a 3-dimensional camera system

In this study a 3-dimensional (3D) camera system was set up to measure individual feed intake of dairy cows in a commercial in-house setting. The system was developed to identify the cows while eating, predict body weight based on the curvature of the back of the cow, and quantify the amount of feed eaten by the cow at each visit of eating. The identification of the cow was based on recognizing the patterns, colors, and curvatures of the back from a reference database obtained in a corridor after milking, where images were taken of all cows with a simultaneous reading of the electronic ear tag. Body weight is predicted using the curvatures of the back of the cow. Feed intake is quantified as the difference in surface of the feed a cow can reach before and after a visit is initiated. This estimate is in liters but converted to kilograms, using the density of the feed in the specific herd. A total of 9,142 cows were measured in 19 herds across 3 breeds: Jersey (2,513 cows), Red Dairy Cattle (2,813 cows), and Holstein (3,816 cows). Mean daily feed intake was higher for Red Dairy Cattle (61.72 kg) and Holstein (64.59 kg) than for Jersey (55.74 kg). Repeatability estimates for daily feed intake as a weekly average was 0.62, 0.65, and 0.63 for Jersey, Red Dairy, and Holstein cattle, respectively. Mean body weight was higher for Red Dairy (647.9 kg) and Holstein (683.8 kg) than for Jersey (469.6 kg). Repeatability estimates for body weight as a weekly average was 0.83, 0.85, and 0.88 for Jersey, Red Dairy, and Holstein, respectively. The perspectives in having such records available is huge both for the farmer and for the dairy industry. The records can both be used for improving management in farms on an individual cow level and herd level, but also for genetic evaluation and selection as well as testing feeding regimens. Feed intake can be measured on an individual level using a 3D camera system.

Genetic parameters for feed intake and body weight in dairy cattle using high-throughput 3-dimensional cameras in Danish commercial farms

Recording complex phenotypes on a large scale is becoming possible with the incorporation of recently developed new technologies. One of these new technologies is the use of 3-dimensional (3D) cameras on commercial farms to measure feed intake and body weight (BW) daily. Residual feed intake (RFI) has been proposed as a proxy for feed efficiency in several species, including cattle, pigs, and poultry. Dry matter intake (DMI) and BW records are required to calculate RFI, and the use of this new technology will help increase the number of individual records more efficiently. The aim of this study was to estimate genetic parameters (including genetic correlations) for DMI and BW obtained by 3D cameras from 6,000 cows in commercial farms from the breeds Danish Holstein, Jersey, and Nordic Red. Additionally, heritabilities per parity and genetic correlations among parities were estimated for DMI and BW in the 3 breeds. Data included 158,000 weekly records of DMI and BW obtained between 2019 and 2022 on 17 commercial farms. Estimated heritability for DMI ranged from 0.17 to 0.25, whereas for BW they ranged from 0.44 to 0.58. The genetic correlations between DMI and BW were moderately positive (0.58-0.65). Genetic correlations among parities in both traits were highly correlated in the 3 breeds, except for DMI between first parity and late parities in Holstein where they were down to 0.62. Based on these results, we conclude that DMI and BW phenotypes measured by 3D cameras are heritable for the 3 dairy breeds and their heritabilities are comparable to those obtained by traditional methods (scales and feed bins). The high heritabilities and correlations of 3D measurements with the true trait in previous studies demonstrate the potential of this new technology for measuring feed intake and BW in real time. In conclusion, 3D camera technology has the potential to become a valuable tool for automatic and continuous recording of feed intake and BW on commercial farms.

Prediction of body condition in Jersey dairy cattle from 3D-images using machine learning techniques

The body condition of dairy cows is a crucial health and welfare indicator that is widely acknowledged. Dairy herds with a well-management body condition tend to have more fertile and functional cows. Therefore, routine recording of high-quality body condition phenotypes is required. Automated prediction of body condition from 3D images can be a cost-effective approach to current manual recording by technicians. Using 3D-images, we aimed to build a reliable prediction model of body condition for Jersey cows. The dataset consisted of 808 individual Jersey cows with 2,253 phenotypes from three herds in Denmark. Body condition was scored on a 1 to 9 scale and transformed into a 1 to 5 scale with 0.5-unit differences. The cows' back images were recorded using a 3D camera (Microsoft Xbox One Kinect v2). We used contour and back height features from 3D-images as predictors, together with class predictors (evaluator, herd, evaluation round, parity, lactation week). The performance of machine learning algorithms was assessed using H2O AutoML algorithm (h2o.ai). Based on outputs from AutoML, DeepLearning (DL; multi-layer feedforward artificial neural network) and Gradient Boosting Machine (GBM) algorithms were implemented for classification and regression tasks and compared on prediction accuracy. In addition, we compared the Partial Least Square (PLS) method for regression. The training and validation data were divided either through a random 7:3 split for 10 replicates or by allocating two herds for training and one herd for validation. The accuracy of classification models showed the DL algorithm performed better than the GBM algorithm. The DL model achieved a mean accuracy of 48.1% on the exact phenotype and 93.5% accuracy with a 0.5-unit deviation. The performances of PLS and DL regression methods were comparable, with mean coefficient of determination of 0.67 and 0.66, respectively. When we used data from two herds for training and the third herd as validation, we observed a slightly decreased prediction accuracy compared to the 7:3 split of the dataset. The accuracies for DL and PLS in the herd validation scenario were > 38% on the exact phenotype and > 87% accuracy with 0.5-unit deviation. This study demonstrates the feasibility of a reliable body condition prediction model in Jersey cows using 3D-images. The approach developed can be used for reliable and frequent prediction of cows' body condition to improve dairy farm management and genetic evaluations.