limap.line2d.SOLD2.model.nets package
Submodules
limap.line2d.SOLD2.model.nets.backbone module
- class limap.line2d.SOLD2.model.nets.backbone.HourglassBackbone(input_channel=1, depth=4, num_stacks=2, num_blocks=1, num_classes=5)
Bases:
Module
Hourglass backbone.
- forward(input_images)
Defines 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.
- training: bool
- class limap.line2d.SOLD2.model.nets.backbone.SuperpointBackbone
Bases:
Module
SuperPoint backbone.
- forward(input_images)
Defines 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.
- training: bool
limap.line2d.SOLD2.model.nets.descriptor_decoder module
- class limap.line2d.SOLD2.model.nets.descriptor_decoder.SuperpointDescriptor(input_feat_dim=128)
Bases:
Module
Descriptor decoder based on the SuperPoint arcihtecture.
- forward(input_features)
Defines 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.
- training: bool
limap.line2d.SOLD2.model.nets.heatmap_decoder module
- class limap.line2d.SOLD2.model.nets.heatmap_decoder.PixelShuffleDecoder(input_feat_dim=128, num_upsample=2, output_channel=2)
Bases:
Module
Pixel shuffle decoder.
- forward(input_features)
Defines 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.
- get_channel_conf(num_upsample)
- training: bool
limap.line2d.SOLD2.model.nets.junction_decoder module
- class limap.line2d.SOLD2.model.nets.junction_decoder.SuperpointDecoder(input_feat_dim=128, backbone_name='lcnn')
Bases:
Module
Junction decoder based on the SuperPoint architecture.
- forward(input_features)
Defines 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.
- training: bool
limap.line2d.SOLD2.model.nets.lcnn_hourglass module
Hourglass network, taken from https://github.com/zhou13/lcnn
- class limap.line2d.SOLD2.model.nets.lcnn_hourglass.HourglassNet(block, head, depth, num_stacks, num_blocks, num_classes, input_channels)
Bases:
Module
Hourglass model from Newell et al ECCV 2016
- forward(x)
Defines 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.
- training: bool
- limap.line2d.SOLD2.model.nets.lcnn_hourglass.hg(**kwargs)