Overview of the modules interactionsLink
The following figure shows a use case of the ROMI modules, and the way they interact, to design an efficient plant phenotyping platform used in research.
Should be totally independent of the rest since it could be uses in other parts of the ROMI project (Rover, Cable bot, ...) through the abstract class
DB or even the local database class
It requires a physical connection to the hardware (
pyserial) to control. It also needs an active ROMI database to export acquired datasets (plant images).
Virtual Plant ImagerLink
It requires a connection to an active ROMI database to export generated datasets (virtual plant images). In case of machine learning methods, a database would also provide training datasets.
Plant 3D VisionLink
It requires connection to an active ROMI database to import datasets to process and export the results. Two plant reconstruction approaches are available in the SmartInterpreter:
- Geometry based, try to infer the plant's geometry using structure from motion algorithms and space carving to first reconstruct a point cloud.
- Machine learning based, try to infer the plant's geometry using semantic (organ) segmentation of pictures and space carving to first reconstruct a labelled point cloud.
Then meshing and skeletonization finally enables to extract the plant's phyllotaxis.
Plant 3D ExplorerLink
It requires a database with datasets to browse and represent.
Research oriented user storyLink
- The user put his/her plant inside the scanner and run acquisitions, which returns a set of images per plant.
- These images are uploaded to a central database.
- The user defines a pipeline to reconstruct and quantify plants architecture by choosing among a set of predefined methods and algorithms. These instructions may be run by a distant server.
- Finally, the user can access the acquisitions, reconstructions & quantitative data by connecting to a visualization server using his/her computer