Deep learning is a central tool in plant phenotyping. Proficiency in artificial intelligence and programming remains essential for designing a deep learning model. The reuse of deep models can be challenging for non-coding end-user such as plant biologists. Currently, very few tools facilitate collaboration between computational infrastructure, non-coding end-users, and deep learning project manager. Consequently, this limits the deployment of multi centric large scale deep learning initiatives (such as the notable exception of the Globally Wheat Challenge) in plant phenotyping. We propose the scheme of Fig. 1 to allow non-coding end-users to test existing deep learning models with a homemade software capable of running any segmentation, classification or object detection deep learning model. This software is an open source plugin coined MANINI (https://github.com/hereariim/manini) which runs under the north American initiative NAPARI (https://napari.org/). The plugin also enables the manual correction of the inference. The corrected inferences can then be used to retrain or fine-tune the model on a large-scale infrastructure (in our case the European Grid Inrastructure EGI) via the DEEPaaS API. The resulting new model is then accessible on a public repository. We illustrate the interest of this scheme on various plant phenotyping use cases.

We investigate the robustness of 3D point-based deep learning for organ segmentation of 3D plant models against varying reconstruction quality of the surface. The reconstruction quality is quantified in two ways: 1) The number of acquisitions for partial 3D scans and 2) the amount of noise. High quality models of real rosebush plants are used to collect point clouds in a controlled simulation environment as a way to degrade surface quality systematically. We show that the well-known 3D point-based neural network PointNet++ is capable of operating effectively on low quality and corrupted data for the task of plant organ segmentation. The results indicate that investing on developing deep learning methods has the potential of advancing applications of automated phenotyping, especially for low-quality 3D point clouds of plants. Keywords: plant phenotyping, organ segmentation, robustness analysis, point-based deep learning (a) (b) Figure 1: A 3D rosebush model from ROSE-X data set: (a) point cloud; (b) triangular mesh model.