Skip to content

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]:
that mean you need to specify the environment variable 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.