Evaluating Neural Radiance Fields for 3D PlantGeometry Reconstruction in Field Conditions

We evaluate different Neural Radiance Field (NeRF) techniques for the 3D reconstruction of plants invaried environments, from indoor settings to outdoor fields. Traditional methods usually fail to capture thecomplex geometric details of plants, which is crucial for phenotyping and breeding studies. We evaluatethe reconstruction fidelity of NeRFs in 3 scenarios with increasing complexity and compare the resultswith the point cloud obtained using light detection and ranging as ground truth. In the most realistic fieldscenario, the NeRF models achieve a 74.6% F1 score after 30 min of training on the graphics processingunit, highlighting the efficacy of NeRFs for 3D reconstruction in challenging environments. Additionally, wepropose an early stopping technique for NeRF training that almost halves the training time while achievingonly a reduction of 7.4% in the average F1 score. This optimization process substantially enhances thespeed and efficiency of 3D reconstruction using NeRFs. Our findings demonstrate the potential of NeRFsin detailed and realistic 3D plant reconstruction and suggest practical approaches for enhancing thespeed and efficiency of NeRFs in the 3D reconstruction process.