Answers are grounded in this paper and cite their sources.
Accurate live body weight (LBW) monitoring is one of the most important operational tasks in precision yak farming, but conventional scale-based weighing is labor-intensive and often impractical in plateau farms. This study proposed a contactless, edge-deployable yak LBW estimation system using dual-view passive stereo vision. Under field imaging conditions, stereo-depth voids and landmark ambiguity can affect morphometric extraction. This study therefore designed a framework integrating a lightweight keypoint localization model using side- and dorsal-view images to extract five yak morphometric traits through stereo depth reconstruction. Developed with 3,039 annotated in-situ frames, the system was evaluated for LBW prediction on 87 yak–date morphometric–weight pairs from 29 yaks. For the two dorsal-view traits, the two-dimensional (2D) planar projection improved downstream GPR-based prediction under a regression-stage animal-disjoint 5-fold cross-validation, increasing the coefficient of determination (R²) from 0.74 to 0.89 and reducing the root mean squared error (RMSE) from 22.77 to 15.06 kg compared with direct three-dimensional (3D) distance-based extraction. Based on the extracted traits, a Gaussian process regression (GPR) model predicted the LBW with an R² of 0.965, an RMSE of 8.33 kg and a mean absolute percentage error (MAPE) of 2.60%. These results indicate that combining lightweight stereo vision with dual-view morphometrics offers a promising proof-of-concept for yak LBW estimation under the present animal-disjoint regression protocol, while broader validation across multiple farms, breeds, sexes, ages and seasons remains necessary.
No takes yet. Share an insight, caveat, or question.
Yao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: