Evaluation of 3D Gaussian Splatting in Plant Reconstruction. McAfee, A., Pluck, T., Dahyot, R., & Lacey, G. In Irish Machine Vision and Image Processing (IMVIP 2025), pages 26-29, Ulster University, Derry-Londonderry, Northern Ireland, 2025.
Evaluation of 3D Gaussian Splatting in Plant Reconstruction [pdf]Paper  Evaluation of 3D Gaussian Splatting in Plant Reconstruction [link]Code  doi  abstract   bibtex   3 downloads  
Accurate 3D reconstruction of plants is essential for applications in precision agriculture, phenotyping, and plant health monitoring. Traditional methods such as LiDAR or Structure-from-Motion often struggle with complex plant topologies or require complex sensing hardware. Recent advances in neural rendering, particularly 3D Gaussian Splatting (3DGS), have shown promise in efficiently capturing fine-grained plant details. In this paper, we evaluate the performance of 3DGS on a new dataset of seven plants, each comprising approximately 500 multi-view images. Results demonstrate that 3DGS effectively reconstructs complex plant structures and is a suitable visual aid for plant phenotyping experts.

Downloads: 3