Body Condition Score Estimation Based on Regression Analysis Using a 3D Camera

Leveraging computer vision, large language models, and multimodal machine learning for optimal decision-making in dairy farming

This article explores various applications of artificial intelligence technologies in dairy farming, including the use of computer vision systems (CVS) for animal identification, body condition score (BCS) and body shape analysis, and potential uses of LLMs in the dairy industry. Among recent advancements in precision livestock farming (PLF) tools, CVS have gained popularity as powerful solutions for individual animal monitoring. These systems can capture phenotypes from multiple animals simultaneously using a single device in an automated and non-intrusive manner. To match animals with their corresponding predicted phenotypes, these systems require individual animal identification, which can be achieved through external identification systems or computer vision-based animal identification algorithms. Additionally, modern natural language processing techniques, such as large language models (LLMs), offer opportunities for advanced data integration, including unstructured textual data. Furthermore, we discuss the challenges associated with integrating data from different sources and modalities - such as images, text, and tabular data - into multimodal machine learning systems for phenotype prediction, which also represents a key area of artificial intelligence application. Digital technologies such as CVS and LLMs have the potential to transform dairy farming. CVS can provide individual and objective assessments of animal health, while LLMs can integrate diverse data sources for phenotype prediction. While there is much potential ahead, these technologies offer significant opportunities for advancing animal health monitoring, farm management, and individual phenotyping.

Monitoring Pig Structural Soundness and Body Weight in Pork Production Systems Using Computer Vision Approaches

As the global demand for products from food-producing animals increases with greater household economic capacity, there is an increased emphasis on the development of precision technologies for monitoring the health, product production, and wellbeing of these animals. The present review focuses on pork production. Using these systems is advantageous for enhancing pork production efficiency when trained personnel utilize these technologies to full capacity and have objective, automated, and uninterrupted streams of data collection. While these systems have great potential for revolutionizing food animal production, the nascent stage of computer vision in precision technology has precluded its integration into traditional agricultural practices and systems. In this review paper, there is a focus on the need to (1) evaluate the performance and effective use of computer vision technologies to collect and evaluate reliable data from pork production enterprises; and (2) focus on the current state of sensor-based animal management using a data fusion approach to monitor pig health/performance. Many of these technologies are in various stages of development; therefore, these technologies have not been integrated into pork production or other food animal producing systems. Even though the focus of this review article is on the utilization of these technologies in pork production systems, these technologies are relevant in other food animal production systems, particularly dairy and poultry production. Therefore, we describe an approach that emphasizes the important need for computational capacity and speed, edge computing, data storage and transmission, and maintaining connectivity in rural settings.

Fertilization rate of crossbreeding cattle using sexing and conventional semen in different seasons in South Papua

Objective

Fertilization rate of artificially inseminated cows using sexed and conventional semen in different seasons in South Papua.

Materials and Methods

Eighty crossbred cows aged 4-4.5 years with body condition score 3.8 were divided into groups A (summer = 40 cows) and B (rainy season = 40 cows). Each cow in each season was artificial insemination (AI) using sexed frozen semen and conventional semen. Frozen semen was evaluated for post-thawing motility (PTM), cell membrane integrity, and acrosome damage before synchronization using 5 ml PGF2α plus vitamin E. Using a visual gun, we identified cows in estrus on days 4-7 post-synchronization. Pregnancy of cows was detected using N5Vet ultrasound on days 35 and 55. The interaction between season, semen type, and fertilization level was analyzed using standard error and two-way ANOVA, assisted by SPSS 21 software.

Results

The wet season Temperature-Humidity Index (THI) level averaged 77.12 ± 1.19, and the summer season THI level averaged 82.67 ± 1.25. PTM quality averaged 60%-65%, viability 61%-71%, sperm membrane integrity 62%-65%, and acrosome integrity 88%-91%. Conception rates (CR) value of rainy season (p < 0.05) with summer season. In addition, the services per conception (S/C) value in the rainy season (p > 0.05) is the same as in the summer. This study's S/C and CR values were within normal limits, and the pregnancy rate reached 65%-86%. Pregnancy detection can be observed on day 35, and the fetal heartbeat is visible.

Conclusion

Post-AI fertilization using conventional semen was better in all seasons. The double dose of sexed semen can increase the fertilization rate in summer.

Comparison of manual and automated body condition scoring of commercial dairy cattle

BACKGROUND

Body condition scoring of dairy cows estimates their body reserves. Automation allows increased data availability and reduced labour costs. The aim of this study was to compare an automated (AUT) body condition score (BCS) system to manual observers on a single commercial dairy farm in south-west England.

METHODS

Three practising veterinary surgeons performed body condition scoring of 315 dairy cows using the agriculture and horticulture development board (AHDB) Body Condition Scorecard. AUT BCSs were obtained from two 3D cameras and compared to the BCSs recorded by the three operators.

RESULTS

The AUT system only agreed with manual scorers at a BCS of 3. The system failed to detect cows classified as underconditioned (BCS ≤ 2.25) by any of the operators (sensitivity 0%). It also systematically underestimated the BCS of cows classified as overconditioned (BCS ≥ 3.5) by the operators. For overconditioned cows, the sensitivity of the AUT system ranged from 30.7% to 48.8% when compared with the manual operators. The AUT system also had weaker agreement with operators for Jersey cows, with Cohen's weighted kappa values of 0.28 for Jersey animals and 0.40 for Holsteins.

LIMITATIONS

This study used a convenience sample of animals on a single farm at a single time point, so the extent to which the findings can be more widely generalised is unclear.

CONCLUSIONS

The AUT system failed to detect animals classified as underconditioned by the operators and underestimated the condition of cows classified as overconditioned by the operators. Currently, without improvements to the algorithm, the clinical usefulness of such an AUT system for body condition scoring is limited.

Automated Cow Body Condition Scoring Using Multiple 3D Cameras and Convolutional Neural Networks

Body condition scoring is an objective scoring method used to evaluate the health of a cow by determining the amount of subcutaneous fat in a cow. Automated body condition scoring is becoming vital to large commercial dairy farms as it helps farmers score their cows more often and more consistently compared to manual scoring. A common approach to automated body condition scoring is to utilise a CNN-based model trained with data from a depth camera. The approaches presented in this paper make use of three depth cameras placed at different positions near the rear of a cow to train three independent CNNs. Ensemble modelling is used to combine the estimations of the three individual CNN models. The paper aims to test the performance impact of using ensemble modelling with the data from three separate depth cameras. The paper also looks at which of these three cameras and combinations thereof provide a good balance between computational cost and performance. The results of this study show that utilising the data from three depth cameras to train three separate models merged through ensemble modelling yields significantly improved automated body condition scoring accuracy compared to a single-depth camera and CNN model approach. This paper also explored the real-world performance of these models on embedded platforms by comparing the computational cost to the performance of the various models.

Opportunities for Regulatory Authorities to Assess Animal-Based Measures at the Slaughterhouse Using Sensor Technology and Artificial Intelligence: A Review

Animal-based measures (ABMs) are the preferred way to assess animal welfare. However, manual scoring of ABMs is very time-consuming during the meat inspection. Automatic scoring by using sensor technology and artificial intelligence (AI) may bring a solution. Based on review papers an overview was made of ABMs recorded at the slaughterhouse for poultry, pigs and cattle and applications of sensor technology to measure the identified ABMs. Also, relevant legislation and work instructions of the Dutch Regulatory Authority (RA) were scanned on applied ABMs. Applications of sensor technology in a research setting, on farm or at the slaughterhouse were reported for 10 of the 37 ABMs identified for poultry, 4 of 32 for cattle and 13 of 41 for pigs. Several applications are related to aspects of meat inspection. However, by European law meat inspection must be performed by an official veterinarian, although there are exceptions for the post mortem inspection of poultry. The examples in this study show that there are opportunities for using sensor technology by the RA to support the inspection and to give more insight into animal welfare risks. The lack of external validation for multiple commercially available systems is a point of attention.

Precision Livestock Farming: What Does It Contain and What Are the Perspectives?

Precision Livestock Farming (PLF) describes the combined use of sensor technology, the related algorithms, interfaces, and applications in animal husbandry. PLF technology is used in all animal production systems and most extensively described in dairy farming. PLF is developing rapidly and is moving beyond health alarms towards an integrated decision-making system. It includes animal sensor and production data but also external data. Various applications have been proposed or are available commercially, only a part of which has been evaluated scientifically; the actual impact on animal health, production and welfare therefore remains largely unknown. Although some technology has been widely implemented (e.g., estrus detection and calving detection), other systems are adopted more slowly. PLF offers opportunities for the dairy sector through early disease detection, capturing animal-related information more objectively and consistently, predicting risks for animal health and welfare, increasing the efficiency of animal production and objectively determining animal affective states. Risks of increasing PLF usage include the dependency on the technology, changes in the human-animal relationship and changes in the public perception of dairy farming. Veterinarians will be highly affected by PLF in their professional life; they nevertheless must adapt to this and play an active role in further development of technology.

Invited review: Assessment of body condition score and body fat reserves in relation to insulin sensitivity and metabolic phenotyping in dairy cows

The purpose of this article is to review body condition scoring and the role of body fat reserves in relation to insulin sensitivity and metabolic phenotyping. This article summarizes body condition scoring assessment methods and the differences between subcutaneous and visceral fat depots in dairy cows. The mass of subcutaneous and visceral adipose tissue (AT) changes significantly during the transition period; however, metabolism and intensity of lipolysis differ between subcutaneous and visceral AT depots of dairy cows. The majority of studies on AT have focused on subcutaneous AT, and few have explored visceral AT using noninvasive methods. In this systematic review, we summarize the relationship between body fat reserves and insulin sensitivity and integrate omics research (e.g., metabolomics, proteomics, lipidomics) for metabolic phenotyping of cows, particularly overconditioned cows. Several studies have shown that AT insulin resistance develops during the prepartum period, especially in overconditioned cows. We discuss the role of AT lipolysis, fatty acid oxidation, mitochondrial function, acylcarnitines, and lipid insulin antagonists, including ceramide and glycerophospholipids, in cows with different body condition scoring. Nonoptimal body conditions (under- or overconditioned cows) exhibit marked abnormalities in metabolic and endocrine function. Overall, reducing the number of cows with nonoptimal body conditions in herds seems to be the most practical solution to improve profitability, and dairy farmers should adjust their management practices accordingly.

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.

Associations of automated body condition scores at dry-off and through early lactation with milk yield of Holstein cows

The objective of this study was to analyze the associations of body condition score (BCS) and BCS change (∆BCS) during the dry period and the first 100 d of lactation with daily milk yield. Examining the involvement of health status in the associations between BCS and milk yield was a secondary objective of this research. Data included 12,042 lactations in 7,626 Holstein cows calving between April 2019 and January 2022 in a commercial dairy operation located in Colorado, USA. BCSs were generated daily by an automated BCS camera system located at the exit of the milking parlor. The assessment points selected for this study were dry-off (BCSdry), calving (BCS1), 7 DIM (BCS7), 14 DIM (BCS14), 21 DIM (BCS21), and nadir (nBCS; defined as the lowest daily BCS from calving to 100 DIM). Subsequently, these BCS were categorized considering quartiles (Q1 = 25% lowest BCS; Q4 = 25% greatest BCS), separately for primiparous and multiparous cows. Changes in BCS were calculated from dry-off to calving (multiparous); and from calving to 7 DIM, 14 DIM, 21 DIM, and nadir and assigned into quartile categories considering Q1 as the 25% of cows with the greatest decrease of BCS. Lactations were classified based on the number of health events before nadir as healthy, affected by one event, or having multiple events. Data were examined in primiparous and multiparous cows separately using ANOVA. The least square means for daily milk at 60 DIM and 305 DIM were calculated by category of BCS and ∆BCS at multiple time points and time periods. Subsequently, lactation curves were created by BCS and ∆BCS categories and by health status. Multivariable models included calving season and BCS1 as covariables. The largest differences in milk yield among categories of BCS and ∆BCS were identified for BCS originated at nadir and for the ∆BCS between calving and nadir. The differences in average daily milk yield between cows in the lowest and the greatest nBCS category (Q1 vs. Q4) were 3.3 kg/d (60 DIM) and 3.4 kg/d (305 DIM) for primiparous cows and 2.4 kg/d (60 DIM) and 2.1 kg/d (305 DIM) for multiparous cows. During the period from calving to nadir, primiparous cows in Q1 (greatest decrease of BCS) produced 4.3 kg/d (60 DIM) and 3.8 kg/d (305 DIM) more than cows in Q4. For multiparous cows, the differences were 3.0 kg/d (60 DIM) and 1.9 kg/d (305 DIM) in favor of Q1 cows. Overall, the associations between BCS and ∆BCS categories and milk yield were not consistent across time and they depended on the parity category. Nonetheless, as the assessment of BCS and ∆BCS approached the nadir, the association between greater milk yield and lower BCS or greater reduction in BCS became more evident.

Dynamics of automatically generated body condition scores during early lactation and pregnancy at first artificial insemination of Holstein cows

The objective of this study was to characterize the association between body condition score (BCS) and BCS change (ΔBCS), determined by an automated camera system at multiple time points, and the subsequent pregnancy per first artificial insemination (P/AI1) of Holstein cows. A retrospective observational study was completed using data collected from 11,393 lactations in 7,928 Holstein cows calving between April 2019 and March 2021 in a commercial dairy operation located in Colorado. Cows were classified as primiparous or multiparous. Scores generated by BCS cameras at dry-off, calving, 21 days in milk (DIM), 56 DIM, and first artificial insemination were selected for the analyses and subsequently categorized as low (≤lower quartile), moderate (interquartile range), and high (≥upper quartile). Changes in BCS were calculated by periods of interest as change from dry-off to calving (multiparous cows); change from calving to 21 DIM; change from calving to 56 DIM; and change from calving to first artificial insemination and assigned into categories as large loss of BCS (top 25% of cows losing BCS); moderate loss (bottom 75% of cows losing BCS); no change (ΔBCS = 0); or gain of BCS (ΔBCS > 0). Data were examined in primiparous and multiparous cows separately using logistic regression and time-to-event analyses. Initial univariable models were followed by multivariable models that considered calving season, occurrence of disease, and milk yield up to 60 DIM as covariables. The logistic regression analyses indicated that in both parity groups the associations between BCS category and P/AI1 were more evident at 21 DIM, 56 DIM, and first artificial insemination, with lower odds of P/AI1 in cows in the low BCS category. Likewise, cows with large loss in BCS between calving and 21 DIM, calving and 56 DIM, and calving and first artificial insemination had lower odds of P/AI1 compared with other categories of ΔBCS within the same period of interest. Similarly, survival analyses evidenced that cows in the low BCS category required more time to get pregnant. In agreement, differences in the dynamics of the average daily BCS during the first 90 DIM were evident when cows were grouped by first AI outcome (pregnant vs. nonpregnant) and by their time to pregnancy category (<90 DIM; 91-150 DIM; or >150 DIM), with cows with reduced fertility showing lower BCS up to 90 DIM. Overall, low BCS and more pronounced reductions in BCS occurring closer to first artificial insemination resulted in lower odds of pregnancy per artificial insemination.

Opportunities to Harness High-Throughput and Novel Sensing Phenotypes to Improve Feed Efficiency in Dairy Cattle

Simple Summary

Sensors, routinely collected on-farm tests, and other repeatable, high-throughput measurements can provide novel phenotype information on a frequent basis. Information from these sensors and high-throughput measurements could be harnessed to monitor or predict individual dairy cow feed intake. Predictive algorithms would allow for genetic selection of animals that consume less feed while producing the same amount of milk. Improved monitoring of feed intake could reduce the cost of milk production, improve animal health, and reduce the environmental impact of the dairy industry. Moreover, data from these information sources could aid in animal management (e.g., precision feeding and health detection). In order to implement tools, the relationship of measurements with feed intake needs to be established and prediction equations developed. Lastly, consideration should be given to the frequency of data collection, the need for standardization of data and other potential limitations of tools in the prediction of feed intake. This review summarizes measurements of feed efficiency, factors that may impact the efficiency and feed consumption of an animal, tools that have been researched and new traits that could be utilized for the prediction of feed intake and efficiency, and prediction equations for feed intake and efficiency presented in the literature to date.

Abstract

Feed for dairy cattle has a major impact on profitability and the environmental impact of farms. Sustainable dairy production relies on continued improvement in feed efficiency as a way to reduce costs and nutrient loss from feed. Advances in breeding, feeding and management have led to the dilution of maintenance energy and thus more efficient dairy cattle. Still, many additional opportunities are available to improve individual animal feed efficiency. Sensing technologies such as wearable sensors, image-based and high-throughput phenotyping technologies (e.g., milk testing) are becoming more available on commercial farm. The application of these technologies as indicator traits for feed intake and efficiency related traits would be advantageous to provide additional information to predict and manage feed efficiency. This review focuses on precision livestock technologies and high-throughput phenotyping in use today as well as those that could be developed in the future as possible indicators of feed intake. Several technologies such as milk spectral data, activity, rumen measures, and image-based phenotypes have been associated with feed intake. Future applications will depend on the ability to repeatably measure and calibrate these data across locations, so that they can be integrated for use in predicting and managing feed intake and efficiency on farm.

Precision Technologies to Address Dairy Cattle Welfare: Focus on Lameness, Mastitis and Body Condition

Specific animal-based indicators that can be used to predict animal welfare have been the core of protocols for assessing the welfare of farm animals, such as those produced by the Welfare Quality project. At the same time, the contribution of technological tools for the accurate and real-time assessment of farm animal welfare is also evident. The solutions based on technological tools fit into the precision livestock farming (PLF) concept, which has improved productivity, economic sustainability, and animal welfare in dairy farms. PLF has been adopted recently; nevertheless, the need for technological support on farms is getting more and more attention and has translated into significant scientific contributions in various fields of the dairy industry, but with an emphasis on the health and welfare of the cows. This review aims to present the recent advances of PLF in dairy cow welfare, particularly in the assessment of lameness, mastitis, and body condition, which are among the most relevant animal-based indications for the welfare of cows. Finally, a discussion is presented on the possibility of integrating the information obtained by PLF into a welfare assessment framework.