MIR.models.TransVFA

class MIR.models.TransVFA(*args, **kwargs)[source]

TransVFA model for image registration.

Parameters:

  • configs_sw – Swin Transformer config.

  • configs – VFA config.

  • device – Device to run the model on.

  • swin_type – Transformer type (‘swin’, ‘dswin’, ‘dswinv2’).

  • return_orginal – If True, return composed grids and stats.

  • return_all_flows – If True, return flows for all decoder levels.

  • max_channels – Upper bound on the encoder-side feature widths min(configs.start_channels * 2**(i+1), max_channels); the matching decoder uses configs.max_channels. The resulting widths are printed at construction. Default: 64.

Forward inputs:

sample – Tuple (mov, fix) tensors of shape [B, 1, *spatial].

Forward outputs:

Flow(s) and optional auxiliary outputs depending on flags.

__init__(configs_sw, configs, device, swin_type='swin', return_orginal=False, return_all_flows=False, max_channels=64)[source]

Methods

__init__(configs_sw, configs, device[, ...])

forward(sample)

Run forward registration.