Convolutional neural network (CNN)-based computer vision systems have been increasingly applied in animal farming to improve animal management, but current knowledge, practices, limitations, and solutions of the applications remain to be expanded and explored. The objective of this study is to systematically review applications of CNN-based computer vision systems on animal farming in terms of the five deep learning computer vision tasks: image classification, object detection, semantic/instance segmentation, pose estimation, and tracking. Cattle, sheep/goats, pigs, and poultry were the major farm animal species of concern. In this research, preparations for system development, including camera settings, inclusion of variations for data recordings, choices of graphics processing units, image preprocessing, and data labeling were summarized. CNN architectures were reviewed based on the computer vision tasks in animal farming. Strategies of algorithm development included distribution of development data, data augmentation, hyperparameter tuning, and selection of evaluation metrics. Judgment of model performance and performance based on architectures were discussed. Besides practices in optimizing CNN-based computer vision systems, system applications were also organized based on year, country, animal species, and purposes. Finally, recommendations on future research were provided to develop and improve CNN-based computer vision systems for improved welfare, environment, engineering, genetics, and management of farm animals.

The unprecedented 2015 outbreaks of highly pathogenic avian influenza (HPAI) H5N2 in the U.S. devastated its poultry industry and resulted in over $3 billion economic impacts. Today HPAI continues eroding poultry operations and disrupting animal protein supply chains around the world. Anecdotal evidence in 2015 suggested that in some cases the AI virus was aerially introduced into poultry houses, as abnormal bird mortality started near air inlets of the infected houses. This study modeled air movement trajectories and virus concentrations that were used to assess the probability or risk of airborne transmission for the 77 HPAI cases in Iowa. The results show that majority of the positive cases in Iowa might have received airborne virus, carried by fine particulate matter, from infected farms within the state (i.e., intrastate) and infected farms from the neighboring states (i.e., interstate). The modeled airborne virus concentrations at the Iowa recipient sites never exceeded the minimal infective doses for poultry; however, the continuous exposure might have increased airborne infection risks. In the worst-case scenario (i.e., maximum virus shedding rate, highest emission rate, and longest half-life), 33 Iowa cases had > 10% (three cases > 50%) infection probability, indicating a medium to high risk of airborne transmission for these cases. Probability of airborne HPAI infection could be affected by farm type, flock size, and distance to previously infected farms; and more importantly, it can be markedly reduced by swift depopulation and inlet air filtration. The research results provide insights into the risk of airborne transmission of HPAI virus via fine dust particles and the importance of preventative and containment strategies such as air filtration and quick depopulation of infected flocks.

Understanding broiler behaviours provides important implications for animal well-being and farm management. The objectives of this study were to classify specific broiler behaviours by analysing data from wearable accelerometers using two machine learning models, K-Nearest Neighbour (KNN) and Support Vector Machine (SVM). Lightweight triaxial accelerometers were used to record accelerations of nine 7-week-old broilers at a sampling frequency of 40 Hz. A total of 261.6-min data were labelled for four behaviours - walking, resting, feeding and drinking. Instantaneous motion features including magnitude area, vector magnitude, movement variation, energy, and entropy were extracted and stored in a dataset which was then segmented by one of the six window lengths (1, 3, 5, 7, 10 and 20 s) with 50% overlap between consecutive windows. The mean, variation, SD, minimum and maximum of each instantaneous motion feature and two-way correlations of acceleration data were calculated within each window, yielding a total of 43 statistic features for training and testing of machine learning models. Performance of the models was evaluated using pure behaviour datasets (single behaviour type per dataset) and continuous behaviour datasets (continuous recording that involved multiple behaviour types per dataset). For pure behaviour datasets, both KNN and SVM models showed high sensitivities in classifying broiler resting (87% and 85%, respectively) and walking (99% and 99%, respectively). The accuracies of SVM were higher than KNN in differentiating feeding (88% and 75%, respectively) and drinking (83% and 62%, respectively) behaviours. Sliding window with 1-s length yielded the best performance for classifying continuous behaviour datasets. The performance of classification model generally improved as more birds were included for training. In conclusion, classification of specific broiler behaviours can be achieved by recording bird triaxial accelerations and analysing acceleration data through machine learning. Performances of different machine learning models differ in classifying specific broiler behaviours.

Radio frequency identification (RFID) technology offers a real-time solution to monitor behavioral responses of individual animals to various stimuli, which provides crucial implications on farm management and animal well-being. The objectives of this study were to (1) develop and describe an ultra-high frequency radio frequency identification (UHF-RFID) system for continuously monitoring feeding and drinking behaviors of individual broilers in group settings; and (2) validate the performance of the UHF-RFID system against video analysis in determining the instantaneous bird number (IBN) and time spent (TS) at feeder and drinker. The UHF-RFID system consisted of cable-tie tags, antennas, a reader and a data acquisition (DAQ) system. The antennas generated electromagnetic fields where tags were detected and registered by the DAQ system. Electromagnetic fields of the antennas were modified to cover areas of concern (i.e. tube feeders and nipple drinkers) through a series of system evaluations and customizations including tag sensitivity test, power adjustment, radio wave shielding, and assessment of interference by add-ons (e.g. plastic wraps for protecting antennas and an empty carton box for zoning out broilers) and feed/feeder. System validation was performed in two experimental rooms, each with 60 tagged broilers. The results showed that the max reading distances of tags with an identical manufacturer’s specification were markedly different, indicating large variations in sensitivity among the tags. Desired electromagnetic fields could be achieved by adjusting the power supplied to antennas and by partially shielding antennas with customized stainless steel sheets. The protection materials and fully loaded feeder had little effect on electromagnetic fields of the antennas. The accuracies of the UHF-RFID system for determining IBN and TS were, respectively, 92.5±4.2% and 99.0±1.2% by the feeder antennas and 94.7±4.2% and 93.7±6.9% by the drinker antennas. It is concluded that the UHF-RIFD system can accurately detect and record feeding and drinking behaviors of individual broilers in group settings and thus is a useful tool for investigating impacts of resource allocations and management practices on these behaviors.

There is a lack of precision tools for automated poultry preening monitoring. The objective of this study was to develop poultry preening behavior detectors using mask R-CNN. Thirty 38-week brown hens were kept in an experimental pen. A surveillance system was installed above the pen to record images for developing the behavior detectors. The results show that the mask R-CNN had 87.2 ± 1.0% MIOU, 85.1 ± 2.8% precision, 88.1 ± 3.1% recall, 95.8 ± 1.0% specificity, 94.2 ± 0.6% accuracy, 86.5 ± 1.3% F1 score, 84.3 ± 2.8% average precision and 380.1 ± 13.6 ms·image⁻¹ processing speed. The six ResNets (ResNet18-ResNet1000) had disadvantages and advantages in different aspects of detection performance. Training parts of the complex network and transferring some pre-trained weights from the detectors pre-trained in other datasets can save training time but did not compromise detection performance and various datasets can result in different transfer learning efficiencies. Resizing and padding input images to different sizes did not affect detection performance of the detectors. The detectors performed similarly within 100–500 region proposals. Temporal and spatial preening behaviors of individual hens were characterized using the trained detector. In sum, the mask R-CNN preening behavior detector could be a useful tool to automatically identify preening behaviors of individual hens in group settings.

Stretching behavior is one of the broiler comfort behaviors that could be used for animal welfare assessment. However, there is currently no methodology for automatic monitoring of stretching behavior under representative production practices. The objectives of this study were to (1) develop a faster region-based convolutional neural network (faster R-CNN) stretching behavior detector for broiler stretching behavior detection, (2) evaluate broiler stretching behaviors under stocking densities (SDs) of 27 (27SD), 29 (29SD), 33 (33SD), and 39 kg/m2 (39SD) and at weeks 4 and 5 of bird ages, and (3) examine the temporal and spatial distribution of broiler stretching behaviors. The results show that the precision, recall, specificity, and accuracy were over 86% on broiler stretching detection across all SDs and bird ages using the faster R-CNN stretching behavior detector. Broilers spent 230-533 sec stretching every day and showed more stretching behaviors under the 29SD, 33SD, and 39SD in week 4 and under the 29SD and 33SD in week 5, as compared to other SDs. They performed less stretching in a couple of hours after light ON and before light OFF but preferred to stretch in areas with less traffic and disturbance, that is, along the fences and away from the inspection aisle. It is concluded that the stretching behavior detector had acceptable performance in detecting broiler stretching, thus being a useful tool for broiler stretching detection. Broiler stretching behavior is affected by SD and bird age and shows temporal and spatial variations.