Men's labour migration and attitudes towards women's sexual autonomy: evidence from the National Family Health Survey 2019-21.

The yak is an important indigenous livestock species distributed across the Qinghai–Tibet Plateau and surrounding alpine regions. It has long adapted to extreme environments characterized by hypoxia, low temperature, intense ultraviolet radiation, and seasonal forage fluctuation. As a major source of meat, milk, draft power, and household income in high-altitude pastoral areas, the yak plays an essential role in regional food security and pastoral economic development []. Previous studies have shown that yaks have evolved integrated adaptations involving the respiratory and circulatory systems, energy metabolism, digestive efficiency, and genetic mechanisms []. In addition, research on genomic adaptation and sustainable production systems has highlighted the importance of improving production efficiency, protecting ecological security, and promoting smart farming in yak husbandry []. Therefore, precise image-based measurement, body measurement, body weight estimation, and behavior monitoring are regarded as key components of intelligent yak farming. With the development of computer vision, deep learning, and intelligent sensing, image- and video-based non-contact measurement has become an important approach in precision livestock farming. Visual data from two-dimensional images, depth images, and three-dimensional point clouds have been widely used for body measurement extraction, body weight prediction, conformation assessment, and behavior recognition in livestock, with advantages in measurement efficiency, reduced human disturbance, and continuous monitoring [–]. Substantial progress has been made in body weight estimation, body measurement, and behavior analysis. For example, cloud–edge collaborative methods and YOLO-based frameworks have been applied to yak body weight estimation under practical conditions []. In yaks, YOLO-based body-parameter detection has been used for live body weight estimation []. In cattle, semantic segmentation, keypoint detection, stereo vision, and point-cloud-based approaches have been applied to body measurement and weight prediction, demonstrating the effectiveness of non-contact visual perception for livestock phenotyping [–]. In addition, face recognition, pose estimation, and skeleton-based behavior classification have been introduced for yak feeding monitoring, behavior recognition, and posture analysis in complex environments [–]. At the methodological level, recent segmentation architectures, including efficient Transformer-based networks, encoder-decoder models, pyramid parsing networks, high-resolution representations, promptable foundation models, real-time segmentation networks, and point-cloud segmentation methods, have provided an important technical basis for livestock segmentation, measurement, and visual phenotyping [–,–]. Meanwhile, yak-related phenotypic studies have increasingly adopted deep learning for body weight estimation, feeding behavior monitoring, behavior classification, pose estimation, and non-contact body measurement, indicating the practical need for robust visual perception methods in yak husbandry [,,–,].