Synthetic Cow

Overcoming challenges of data scarcity

This project is a pipeline for creating synthetic data from a 3D artist-created model of a cow to be used for training pose estimation models. This project managed to offer a solution to reduce the need for acquiring real-world data, and increase the efficiency of the models. Right now, the focus of this project is on designing and creating models for detection, tracking, keypoint estimation (Ghassemi et al., 2024), and segmentation of cattles for production level behavioral analysis.

References

2024

  1. Moo-ving Beyond Tradition: Revolutionizing Cattle Behavioural Phenotyping with Pose Estimation Techniques
    Navid Ghassemi, Ali Goldani, Ian Q Whishaw, and 1 more author
    arXiv preprint arXiv:2408.06336, 2024