Advanced configuration#

We have implemented several mechanism to allow users to use CAREamics in contexts we do not explicitly support. In this section, we describe several of these mechanisms.

In the future, we hope to add more depending on user requests.

Custom data type#

The data_type parameter of the DataConfig class is a string that is used to choose the data loader within CAREamics. We currently only support array and tiff explicitly.

However, users can set the data_type to custom and use their own read function.

from careamics.config import DataConfig

data_config = DataConfig(
    data_type="custom",  # (1)!
    axes="YX",
    patch_size=[128, 128],
    batch_size=8,
    num_epochs=20,
)

1. As far as the configuration is concerned, you only set the data_type to custom. The rest happens in the CAREamist instance.

Custom AI model#

CAREamics currently only support UNet models, but users can create their own model and use it in CAREamics. First, the model needs to be registered with the register_model decorator, then both the algorithm of AlgorithmConfig and the architecture of the model need to be set to custom.

from careamics.config import FCNAlgorithmConfig, register_model
from torch import nn, ones


@register_model(name="linear_model")  # (1)!
class LinearModel(nn.Module):
    def __init__(self, in_features, out_features, *args, **kwargs):
        super().__init__()
        self.in_features = in_features
        self.out_features = out_features
        self.weight = nn.Parameter(ones(in_features, out_features))
        self.bias = nn.Parameter(ones(out_features))

    def forward(self, input):
        return (input @ self.weight) + self.bias


config = FCNAlgorithmConfig(
    algorithm_type="fcn",
    algorithm="custom",  # (2)!
    loss="mse",
    model={
        "architecture": "custom",  # (3)!
        "name": "linear_model",  # (4)!
        "in_features": 10,
        "out_features": 5,
    },
)

1. Register your model using the decorator and indicates its name.
2. Set the algorithm to custom.
3. In the model, set the architecture to Custom. Watch the capital letter!
4. Indicate the name of the model.