🥳 Excited to share that our new preprint, “Evaluating Transfer Learning Strategies for Improving Dairy Cattle Body Weight Prediction in Small Farms Using Depth-image and Point-cloud Data,” is now available on arXiv.

Body weight is a key indicator of energy balance, health, and productivity in dairy cattle, yet frequent measurement on farms remains challenging. Computer vision provides a practical, non-invasive approach for automated body weight estimation, but its application on small farms is often limited by the availability of training data.

In this study, we addressed two questions. First, can transfer learning from a large farm improve body weight prediction on small farms with limited data? Second, do depth images and three-dimensional point clouds differ in predictive performance when evaluated under the same conditions? Using data from multiple farms, we found that transfer learning consistently improved prediction accuracy on small farms across all models. We also observed no consistent advantage of depth images over point clouds. Overall, these results suggest that transfer learning is a practical solution for small-farm settings where cross-farm data sharing is constrained by privacy, logistical, or policy considerations, as it requires only pretrained model weights rather than raw data.

For more details, please see our paper at: https://lnkd.in/eUXF6cer