Be The Change (BTC)

class torchgeo.models.BTC(backbone, backbone_pretrained=False, classes=1)

Bases: Module

Be The Change (BTC) change detection model.

If you use this model in your research, please cite the following paper:

• https://arxiv.org/abs/2507.03367

Added in version 0.8.

__init__(backbone, backbone_pretrained=False, classes=1)

Initialise BTC model.

Parameters:

• backbone (str) – backbone type (either swin_tiny, swin_small or swin_base).

• classes (int) – number of classes.

• backbone_pretrained (bool) – whether the cityscapes pretrained swin is used.

forward(x)

BTC forward call.

Extract multi-resolution features, fuse by subtraction, decode with UperNet.

Parameters:

x (Tensor) – input image tensor (b, t*c, h, w)

Returns:

binary change map prediction [b, n_cls, h, w].

Return type:

Tensor