microsim

Microscope simulation library for generating realistic microscope images.

Installation

from PyPI

pip install "microsim[all]"

Note

At the moment, this library is a bit more "application" than it is "library". If you are following the docs or tutorials, its probably best to install the full [all] extra, which brings in io, visualization, and other data-fetching dependencies. However, the bare minimal install is fully functional for generating data.

From github

To get the bleeding edge version, which will change rapidly, you can install from github.

pip install "microsim[all] @ git+https://github.com/tlambert03/microsim"

With GPU support

If available, microsim can use either Jax or Cupy to accelerate computations. These are not installed by default, see the jax or cupy installation instructions, paying attention to your GPU requirements. Support for torch is not fully implemented yet, but is planned.

Quickstart

Construct and run a microsim.Simulation object.

using objectsusing dicts

from microsim import schema as ms
from microsim.util import ortho_plot

sim = ms.Simulation(
    truth_space=ms.ShapeScaleSpace(shape=(128, 512, 512), scale=(0.02, 0.01, 0.01)),
    output_space={'downscale': 8},
    sample=ms.Sample(
        labels=[ms.MatsLines(density=0.5, length=30, azimuth=5, max_r=1)]
    ),
    modality=ms.Confocal(pinhole_au=0.2),
)
result = sim.run()
ortho_plot(result)

If you don't like importing all the names, you can also use dictionaries, and pydantic will convert them to the correct objects.

from microsim import Simulation
from microsim.util import ortho_plot

sim = Simulation(
    truth_space={'shape': (128, 512, 512), 'scale': (0.02, 0.01, 0.01)},
    output_space={'downscale': 8},
    sample=dict(
        labels=[
            {   
                'type': 'matslines',
                'density': 0.5,
                'length': 30,
                'azimuth': 5,
                'max_r': 1,
            }
        ]
    ),
    modality={'type': 'confocal', 'pinhole_au': 0.2}
)
result = sim.run()
ortho_plot(result)

Note that you can also construct a simulation from a JSON object. This is particularly useful with the output parameter, to dump the result of the simulation to a file.

confocal.json

{
    "truth_space": {
        "shape": [128, 512, 512],
        "scale": [0.02, 0.01, 0.01]
    },
    "output_space": {
        "downscale": 8
    },
    "sample": {
        "labels": [
        {
            "type": "matslines",
            "density": 0.5,
            "length": 30,
            "azimuth": 5,
            "max_r": 1
        }
        ]
    },
    "modality": {
        "type": "confocal",
        "pinhole_au": 0.2
    },
    "output": "au02.tiff"
}

... and then load a Simulation from that file and run().

from microsim import Simulation
from pathlib import Path

spec = Path('confocal.json').read_text()
sim = Simulation.model_validate_json(spec)
sim.run()

For more details, see the tutorial, conceptual overview and the API docs