TimeAveragingNet

class deepinv.models.TimeAveragingNet(backbone_net)[source]

Bases: Module, TimeMixin

Time-averaging network wrapper.

Adapts a static image reconstruction network for time-varying inputs to output static reconstructions. Average the data across the time dim before passing into network.

Note

The input physics is assumed to be a temporal physics which produced the temporal measurements y (potentially with temporal mask mask). It must either implement a to_static method to remove the time dimension, or already be a static physics (e.g. deepinv.physics.MRI).

Example:

>>> from deepinv.models import UNet, TimeAveragingNet
>>> model = UNet(scales=2)
>>> model = TimeAveragingNet(model)
>>> y = rand(1, 1, 4, 8, 8) # B,C,T,H,W
>>> x_net = model(y, None)
>>> x_net.shape # B,C,H,W
torch.Size([1, 1, 8, 8])

Parameters:

backbone_net (torch.nn.Module) – Base network which can only take static inputs (B,C,H,W)
device (torch.device) – cpu or gpu.

forward(y, physics: TimeMixin, **kwargs)[source]

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.