The study, led by Yanchao Wang and Kaiying Wang with colleagues at institutions in Zhejiang Province, was published in the journal Smart Agricultural Technology. The team built their system around a modified deep learning architecture they call HSM-YOLO, short for Hu Sheep Mounting-YOLO, named after the Hu sheep breed they studied, which is recognized in China as a national-level protected livestock genetic resource. Their approach addresses a persistent gap in precision livestock farming: while computer vision systems for detecting mounting behavior in cattle and pigs have advanced rapidly, sheep have received comparatively little attention, and existing sheep models tend to be bulky, computationally expensive, and tested only in single, fixed farming environments.

The challenge the researchers set out to solve is harder than it might appear. In dense group-housed environments, sheep exhibit complex social interactions that include not only sexual mounting related to estrus but also non-sexual mounting triggered by food competition, social play, and stress responses. Both behavioral patterns look strikingly similar on camera: one animal places its forelimbs on the back of another, stands on its hind limbs for support, and maintains a forward-leaning posture. The crucial distinction lies in the details. Non-sexual mounting typically lacks the rhythmic pelvic thrusting movements of reproductive mounting and shows arbitrary forelimb positioning, manifesting primarily as simple dorsal pressure. These subtle postural differences are exactly what the new model was designed to learn.

To train and test the system, the researchers collected video from two commercial sheep farms in Huzhou City between April 2024 and February 2025, using healthy Hu sheep aged six to twelve months. Rather than relying on a single camera setup, they deliberately engineered four distinct imaging scenarios to push the model’s adaptability. Five dome cameras installed three meters above ground captured footage of side pens near windows, where fluctuating light, fence bars, and sheep clustering created interference. Nine cameras at 2.75 meters monitored the central pens, which suffered from insufficient illumination, close-range occlusions, and animals temporarily exiting the field of view. Three cameras of a different model provided supplementary imaging, and an inspection robot circling the entire barn at 0.3 meters per second added motion blur to the mix of challenges.

From this footage, the team manually selected 867 video clips ranging from roughly three to twenty seconds, then extracted frames and applied a perceptual hashing algorithm to eliminate redundant near-identical images. The result was a curated dataset of 6,655 images divided into training, validation, and test sets at an approximate six-to-two-to-two ratio. Crucially, the annotation scheme was innovative: rather than treating all mounting as a single category, veterinary-supervised labeling divided the interactions into three classes: sexual mounting, non-sexual mounting, and the mounted ewe herself. Non-sexual mounting was introduced primarily to absorb interference, so labels for mounted individuals were not required for that category. This three-class structure allows the model to simultaneously recognize the behavior and localize the fertile animal.

The technical heart of HSM-YOLO lies in four coordinated modifications to the YOLOv11n architecture, one of the most computationally efficient models in the YOLO family. First, the researchers replaced most standard convolutional layers in the backbone network with depthwise separable convolutions, a design that splits feature extraction into a depthwise operation applying a single filter per input channel and a pointwise operation working across channels. The mathematics is compelling: where a standard convolution requires k²×n×m parameters, the separable version needs only k²×n plus n×m, a compression ratio of one over m plus one over k². This dramatically cuts parameters and computation while preserving feature representation, reserving computational overhead for the more demanding modules that follow.

Second, the team upgraded the C3k2 blocks in the backbone to a novel dual-path module called C3k2_BiRep, which flexibly switches between two sub-architectures depending on configuration. One path employs a RepViTBlock with Vision Transformer characteristics, using depthwise convolution as a token mixer to capture long-range dependencies and positional associations between interacting sheep in overlap scenarios. The other path uses a RepBottleneck with multi-branch RepConv structures that fuse three-by-three convolutions, one-by-one convolutions, and identity branches during training, then collapse them into equivalent single convolutions at inference, enhancing fine-grained posture recognition without added cost. A squeeze-and-excitation module within the block adaptively amplifies key channel responses while suppressing background noise from overlapping animals.

Third, an efficient multi-scale attention module in the network’s neck divides feature maps into groups along the channel dimension without reduction, establishing collaborative attention mappings along both horizontal and vertical directions through parallel branches. Two channel attention maps are aggregated through multiplication to achieve cross-channel feature enhancement, while spatial attention weights derived through global average pooling and Softmax operations help the model focus on the key regions associated with mounting behavior. Fourth, the bounding-box regression loss was replaced with a hybrid Inner-SIoU function that combines the angular cost term of SIoU, which guides optimization for low-quality anchors by minimizing angular deviation between predicted and ground-truth boxes, with auxiliary bounding boxes from the Inner-IoU concept that enhance robustness against the drastic scale variations common in crowded pens.

The performance results demonstrate that these four components work synergistically. Trained from scratch on identical hardware, HSM-YOLO achieved a mean Average Precision of 90.2 percent with an F1 score of 88.1 percent, compared to 84.5 percent mAP for the unmodified YOLOv11n baseline. That 5.7 percentage point improvement came alongside a 15.6 percent reduction in computational complexity, down to 5.4 billion floating-point operations, and a model containing just 2.2 million parameters in a 4.5 megabyte package. The system processed 135.9 frames per second on an RTX 4090 graphics card, outpacing every competing model tested, including RT-DETR-l, YOLOv8n, YOLOv10n, and YOLOv12n. Ablation experiments confirmed each module’s contribution: the dual-path C3k2_BiRep added 3.0 points of mAP, the attention mechanism 2.7 points, and the new loss function 2.2 points, while the separable convolutions cut computation by 18.8 percent.

Perhaps most tellingly, the labeling strategy itself proved decisive. When the team compared their three-class annotation against a simpler two-class approach that treated non-sexual mounting as background, the three-class version improved mAP by 2.9 percentage points. The difference was most dramatic for false positives on the mounted class, which plummeted from 954 to 222, a reduction of 76.7 percent, showing that explicitly teaching the model about non-reproductive mounting prevents it from misclassifying irrelevant sheep instances as fertile animals. Across the four imaging scenarios, HSM-YOLO consistently outperformed competitors, achieving 94.6 percent mAP in the favorable side-pen conditions and maintaining 77.3 percent even under the robot’s dynamic viewpoints and motion blur, where rival models degraded severely.

The practical implications extend beyond benchmark numbers. The researchers built a complete monitoring system with a graphical interface supporting real-time detection from single images, local video files, and RTSP IP camera streams, triggering alerts whenever sexual mounting is detected. Testing with 2560-by-1440 resolution video at 25 frames per second showed an average end-to-end detection speed of 47.9 frames per second, comfortably exceeding the playback rate of the source footage. The team also deployed the model on an NVIDIA Jetson Orin NX edge device, where its TensorRT engine ran 7.4 percent smaller and 12.7 percent faster in latency than the baseline YOLOv11, reaching 15.5 end-to-end frames per second.

Generalization testing on 850 images from an entirely different commercial farm in Xiaoshan District, Hangzhou, with distinct lighting, pen layouts, and camera viewpoints, showed HSM-YOLO maintaining 78.8 percent mAP versus 72.6 percent for the baseline, with smaller performance degradation. The researchers acknowledge limitations: the model was trained exclusively on Hu sheep in Zhejiang Province, cannot identify the individual identity of the mounted ewe, lacks distant small-scale samples, and struggles with continuous real-time analysis on memory-constrained edge hardware. Future work will collect video from multiple breeds including Small-tailed Han and Dorper sheep, introduce infrared imaging for all-weather monitoring, integrate tracking and animal re-identification to build a unified detection-tracking-identification framework, and pursue further lightweighting through channel pruning and knowledge distillation. If successful, these advances could transform estrus detection from a labor-intensive guessing game into an automated, round-the-clock precision operation for the sheep farming industry.

Subject of Research: AI Camera System Spots Fertile Ewes in Seconds During Crowded Barn Mounting Events

Article Title: Detection of mounting behavior and mounted ewes in group-housed environments using HSM-YOLO algorithm

Article References: Wang, Y., Gao, M., Hua, Z., Shakeel, A., Ren, C., & Wang, K. (2026). Detection of mounting behavior and mounted ewes in group-housed environments using HSM-YOLO algorithm. Smart Agricultural Technology, 15, Article 102639. https://doi.org/10.1016/j.atech.2026.102639

Image Credits: AI Generated

DOI: Not provided

Keywords: HSM-YOLO, estrus detection, sheep farming, computer vision, deep learning, YOLOv11n, precision livestock farming, mounting behavior, Hu sheep, agricultural AI, Detection, mounting

The story begins in 2013, when buds of Arundo donax were flown aboard the Shenzhou-10 spacecraft. Exposure to the space environment—cosmic radiation, microgravity, and other stressors—induced genetic and phenotypic changes in the plant material. Back on Earth, researchers propagated the buds through tissue culture and carried out successive rounds of selection, ultimately producing a stable mutant line they call space-mutated Arundo donax, or SA. Giant reed is already an attractive candidate for forage because of its enormous biomass and tolerance for marginal environments, but the wild-type plant has drawbacks: its fiber content is high and it overwinters poorly. The space-mutated line retained the parent plant’s hardiness while showing improved nutritional characteristics, with hay containing 19.12 percent crude protein on a dry-matter basis.

To test whether SA could improve meat quality and not just fill stomachs, the team ran a 67-day feeding trial with 30 three-month-old Hu lambs, a Chinese breed prized for its meat. The lambs were split into three groups of ten. A control group received a standard ration based on alfalfa hay and protein supplements from soybean and cotton meal. A second group had its alfalfa hay entirely replaced with SA hay, while a third group had SA substituted for the protein supplements in its concentrate. All rations met the Chinese national standard for meat-type sheep, and the animals grew normally throughout the trial, confirming that SA is a viable feed ingredient.

After the feeding period, six lambs of similar weight from each group were humanely slaughtered, and samples of the longissimus thoracis muscle—the cut that becomes ribeye and similar steaks—were snap-frozen for analysis. Histological examination revealed that muscle fibers in both SA-fed groups had significantly smaller diameters and cross-sectional areas than those of the control animals, a change generally associated with improved tenderness. More importantly, the intramuscular fat content was significantly higher in both SA groups. Intramuscular fat, known as marbling in beef circles, is a key reservoir of flavor precursors and a major determinant of juiciness and overall eating quality.

The mechanism behind this fat boost appears to lie in the carnitine shuttle, the cellular machinery that ferries long-chain fatty acids into mitochondria, where they are burned for energy. The rate-limiting step is catalyzed by an enzyme called carnitine palmitoyltransferase 1, or CPT1, which sits on the outer mitochondrial membrane. When the researchers measured CPT1 levels, they found them reduced in the SA-fed animals, particularly in the group whose protein supplements had been replaced. Consistent with this, lipidomic analysis showed that long-chain acylcarnitines—molecular signatures of fatty acids being shuttled into mitochondria—were consistently downregulated in SA-fed muscle. Fatty acids that would otherwise have been oxidized were instead redirected into storage as triglycerides and phospholipids.

Crucially, this metabolic rewiring did not come at the cost of oxidative stress. Reactive oxygen species levels were unchanged across all three groups, and malondialdehyde, a marker of lipid peroxidation, was actually significantly lower in the alfalfa-replacement group. That matters because excessive mitochondrial fat oxidation is known to generate reactive oxygen species and promote rancidity. The SA-fed lambs apparently achieved fat deposition without the oxidative damage that often accompanies it, possibly thanks to plant-derived secondary metabolites such as terpenoids, flavonoids, and alkaloids in the mutated grass, compounds known to modulate mitochondrial function and antioxidant defenses.

To see how these lipid changes translated into flavor, the team used headspace solid-phase microextraction coupled with gas chromatography–mass spectrometry to profile the volatile compounds in the meat. They identified 535 volatiles, dominated by aldehydes, ketones, and esters. In the alfalfa-replacement group, lipid oxidation products such as 2,3-octanedione were significantly reduced, while desirable aroma compounds—fruity lactones and esters like 5-heptyldihydro-2(3H)-furanone and 3-methylbutyl 3-methylbutanoate—were significantly increased. Using relative odor activity values to weigh each compound’s sensory contribution, the researchers identified 43 key odorants and found that SA feeding shifted the overall aroma profile away from oily, fatty notes and toward green, sweet, and fruity ones.

The final piece of evidence came from a correlation analysis linking the lipidome to the volatilome. Long-chain acylcarnitines such as carnitine C16:1, C18:1, and C20:2 showed robust positive correlations with lipid oxidation-derived volatiles like nonanal and 2,3-octanedione, the compounds responsible for rancid and greasy aromas. Meanwhile, plant-derived terpenoids such as beta-ionone and farnesene, which the animals presumably picked up from the SA forage, contributed fresh, floral, woody, and herbal nuances that enriched the flavor hierarchy and dampened undesirable gamey odors. Together, these data sketch out what the authors call a carnitine shuttle–lipid remodeling–flavor axis: the forage alters mitochondrial fat transport, which alters fat storage, which alters the chemistry of aroma formation during cooking.

The implications extend beyond a single breed of sheep. Mutton’s distinctive flavor, driven largely by branched-chain fatty acids and lipid-derived volatiles, is a major barrier to consumer acceptance in many markets, so a feed-based strategy for softening it could have real commercial value. Just as significant is the agricultural angle: giant reed grows vigorously on marginal land with few inputs, and a space-mutated line that doubles as nutritious forage could ease pressure on alfalfa and protein concentrate supplies. The study also adds to a growing body of evidence that space mutagenesis is a practical plant-breeding tool, not a novelty. As the authors note, future work with sensory panels, enzyme assays, and sampling across slaughter stages will be needed to confirm the mechanism, but the prospect of space-flown grass improving the taste of Sunday dinner is already on the table.

Subject of Research: Effects of space-mutated Arundo donax forage on carnitine-mediated lipid metabolism and mutton flavor in Hu sheep

Article Title: Dietary substitution with space-mutated Arundo donax L. remodels carnitine-mediated lipid oxidation and improves mutton flavor in Hu sheep

Article References: Nie, Y., Zhang, L., Han, C., Li, X., Wu, Y., Wang, C., & Ma, Y. (2026). Dietary substitution with space-mutated Arundo donax L. remodels carnitine-mediated lipid oxidation and improves mutton flavor in Hu sheep. Food Chemistry: X, Article 104555. https://doi.org/10.1016/j.fochx.2026.104555

Image Credits: AI Generated

DOI: 10.1016/j.fochx.2026.104555

Keywords: space mutagenesis, Arundo donax, Hu sheep, mutton flavor, carnitine shuttle, CPT1, lipidomics, intramuscular fat, volatile compounds, fatty acid oxidation, animal feed, food chemistry

Hu sheep are a distinctive breed, prized for their early sexual maturity, remarkable prolificacy, and ability to thrive in the hot, humid climates of eastern China. They are raised primarily for meat, and traits such as live weight and carcass weight translate directly into farm income. Yet the genetic architecture behind these traits has remained largely opaque. Growth and slaughter characteristics are classic complex traits: they are shaped by thousands of genetic variants scattered across the genome, each contributing a small effect, and by environmental factors ranging from feed quality to husbandry. Untangling which variants matter requires both large sample sizes and statistical methods robust enough to separate true genetic signals from noise.

The research team, led by scientists from Ludong University and Yangzhou University’s Key Laboratory for Animal Genetics and Molecular Breeding of Jiangsu Province, in collaboration with the Chinese Academy of Sciences and a commercial farming partner, designed the study in two stages. A discovery cohort of 112 sheep was used to hunt for initial associations, and an independent validation cohort of 308 animals was then employed to confirm that the signals held up. This two-cohort design is a critical safeguard in genomics research, where spurious associations can arise easily when hundreds of thousands of genetic markers are tested simultaneously against multiple traits.

The genotyping effort was formidable. High-throughput sequencing of the animals’ genomes yielded 559,996 high-quality single nucleotide polymorphisms, or SNPs — positions in the DNA sequence where a single letter of the genetic code varies between individuals. Before any association testing, the team applied stringent quality filters, checking each variant for minor allele frequency and Hardy-Weinberg equilibrium, and used principal component analysis to control for population structure, the subtle genetic substructure within the flock that can masquerade as a trait association. Association tests were run under a mixed linear model, a statistical framework that accounts for relatedness among individuals and has become the gold standard for genome-wide association studies in livestock.

The GWAS surfaced multiple significant loci connected to economically vital traits, including live weight and carcass weight. But the study’s real innovation lay in its second layer: expression quantitative trait locus, or eQTL, analysis. An eQTL is a genetic variant that influences not a visible trait directly, but the activity level — the expression — of a nearby or distant gene. By pairing the SNP genotypes with RNA sequencing data from the longissimus lumborum, the large muscle running along the spine that constitutes much of the prized loin cut, the researchers could ask a more mechanistic question: which DNA variants actually change how genes behave in muscle tissue?

That analysis detected 2,368 cis-eQTLs, meaning variants that regulate the expression of genes located close to them on the same chromosome. Among this entire catalog, the single most significant eQTL in the dataset was linked to the expression of ZNF280B in the longissimus lumborum muscle. This convergence was striking: a gene whose expression levels were more strongly tied to its local genetic background than that of any other gene examined, sitting in a genomic region associated with the very traits breeders care most about. In other words, the statistical map of traits and the molecular map of gene regulation pointed at the same address on the sheep genome.

ZNF280B encodes a zinc-finger protein, a class of DNA-binding molecules that typically function as transcriptional regulators, switching other genes on or off. To determine whether the gene’s correlation with growth traits reflected genuine biological causation rather than mere coincidence, the team turned to functional validation in the laboratory. Using RNA interference, a technique that silences specific genes by degrading their messenger RNA, the researchers reduced ZNF280B activity in skeletal muscle satellite cells — the stem-like cells responsible for muscle growth and repair throughout an animal’s life.

The results were unambiguous. When ZNF280B was downregulated, the satellite cells proliferated significantly more slowly, and the cells showed clear signs of induced apoptosis, the controlled program of cell death. This dual effect — less cell division and more cell death — demonstrates that ZNF280B is not a passive bystander in muscle biology but an active supporter of the cellular processes that build muscle mass. The authors also highlight a role for the gene in fat metabolism, suggesting it may influence the balance between lean tissue and fat deposition that determines carcass quality. A gene that simultaneously governs proliferation, survival, and metabolic fate in muscle cells is exactly the kind of pleiotropic regulator that can leave a measurable imprint on whole-animal traits.

What distinguishes this study from much of the livestock genomics literature is the completeness of the evidence chain. Many GWAS reports stop at statistical associations, leaving the biological mechanism speculative. Here, the pipeline moved from population-level association, through tissue-level gene regulation, down to single-cell-type functional perturbation — a progression that substantially raises confidence that ZNF280B genuinely contributes to variation in growth and carcass traits rather than merely tagging a nearby causal variant. The large sample size of 420 animals, the independent validation cohort, and the multi-trait scope further strengthen the findings, addressing the replication and power problems that have historically plagued association studies in agricultural species.

The practical implications are immediate. Reliable genetic markers linked to ZNF280B expression could be incorporated into marker-assisted selection programs, allowing breeders to identify, early in life, animals genetically predisposed to superior growth and carcass yield — long before those traits can be measured directly. For a breed like Hu sheep, whose adaptability and prolificacy make it a cornerstone of sustainable sheep production in China’s challenging climates, even modest genetic gains compound across generations. More broadly, the study offers a template for livestock genomics: sequence deeply, associate broadly, map expression, and then validate function. As the authors note, the work deepens understanding of the genetic mechanisms underlying growth and slaughter performance and points the way toward production systems that are both more efficient and more sustainable — a goal that matters as global demand for animal protein continues to rise against finite land and feed resources.

Subject of Research: Genetic regulation of growth and carcass traits in Hu sheep through integrative GWAS, eQTL mapping, and functional validation of ZNF280B

Article Title: Integrative GWAS, eQTL, and functional validation identify ZNF280B as a key regulator of growth and carcass traits in Hu sheep

Article References: Wang, Q., Xu, J., Bao, M., Wang, H., Sun, X., Liu, Q., Wang, D., Wang, J., Sun, G., & Li, Y. (2026). Integrative GWAS, eQTL, and functional validation identify ZNF280B as a key regulator of growth and carcass traits in Hu sheep. BMC Genomics. https://doi.org/10.1186/s12864-026-13286-6

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13286-6

Keywords: Hu sheep, GWAS, eQTL, ZNF280B, SNPs, carcass traits, muscle development, marker-assisted selection, livestock genomics, skeletal muscle satellite cells, BMC Genomics, animal breeding