Reconstruct 3D data from imagesLink
Plant reconstruction and analysis pipeline
Getting startedLink
To follows this guide you should have:
- installed the necessary ROMI software here or followed the instructions for the docker image here
- access to a database with a "plant acquisition" to reconstruct (or use the provided examples)
Reconstruction pipelineLink
Cleaning a datasetLink
If you made a mess, had a failure or just want to start fresh with your dataset, no need to save a copy on the side, you can use the Clean
task:
romi_run_task Clean integration_tests/2019-02-01_10-56-33 \
--config plant3dvision/configs/original_pipe_0.toml --local-scheduler
Here the config may use the [Clean]
section where you can define the force
option:
[Clean]
force=true
If true
the Clean
task will run silently, else in interactive mode.
Geometric pipelineLink
Real scan datasetLink
The full geometric pipeline, ie. all the way to angles and internodes measurement, can be called on real dataset with:
romi_run_task AnglesAndInternodes integration_tests/2019-02-01_10-56-33 \
--config plant3dvision/configs/original_pipe_0.toml --local-scheduler
Note
This example uses a real scan dataset from the test database.
Virtual plant datasetLink
The full geometric pipeline, ie. all the way to angles and internodes measurement, can be called on a virtual dataset with:
romi_run_task AnglesAndInternodes integration_tests/arabidopsis_26 \
--config plant3dvision/configs/original_pipe_0.toml --local-scheduler
Note
This example uses a virtual scan dataset from the test database.
Warning
If you get something like this during the Voxel
tasks:
Choose platform:
[0] <pyopencl.Platform 'NVIDIA CUDA' at 0x55d904d5af50>
Choice [0]:
PYOPENCL_CTX='0'
Machine Learning pipelineLink
Warning
This requires the installation of the romiseg
libraries (see here for install instructions) and a trained PyTorch model!
Note
A trained model, to place under <dataset>/models/models
, is accessible here: https://media.romi-project.eu/data/Resnetdataset_gl_png_896_896_epoch50.pt
Real scan datasetLink
The full geometric pipeline, ie. all the way to angles and internodes measurement, can be called on real dataset with:
romi_run_task PointCloud integration_tests/2019-02-01_10-56-33
--config plant3dvision/configs/ml_pipe_vplants_3.toml --local-scheduler
Note
This example uses a real scan dataset from the test database.
Virtual plant datasetLink
The full geometric pipeline, ie. all the way to angles and internodes measurement, can be called on a virtual dataset with:
romi_run_task PointCloud integration_tests/arabidopsis_26 \
--config plant3dvision/configs/ml_pipe_vplants_3.toml --local-scheduler
Note
This example uses a virtual scan dataset from the test database.