PLANesT-3D: A New Annotated Data Set of 3D Color Point Clouds of Plants.

In this paper, we introduce a new data set, named as PLANesT-3D, which includes complete 3D color point clouds of plants. PLANesT-3D is composed of 34 point clouds acquired from three different plant species: Pepper, rosebush and ribes. The point clouds were reconstructed from 2D color images of real plants through multiview stereo. Through a semi-automatic process, background and noise removal was performed. The point clouds were scaled to their true metric dimensions and brought to a particular pose. Each point in the point clouds was manually labeled into "leaf" and "stem" classes. Also, each leaf instance was assigned a leaf identification number. In order to provide benchmark semantic segmentation performance, segmentation results obtained with the application of PointNet++ are reported. We believe that PLANesT-3D data set will contribute to the development and assessment of segmentation methods that operate on 3D color plant point clouds.