MCPcopy Create free account
hub / github.com/baegwangbin/surface_normal_uncertainty / forward

Method forward

models/baseline.py:15–23  ·  view source on GitHub ↗
(self, x, **kwargs)

Source from the content-addressed store, hash-verified

13 self.decoder = Decoder(num_classes=4)
14
15 def forward(self, x, **kwargs):
16 out = self.decoder(self.encoder(x), **kwargs)
17
18 # Bilinearly upsample the output to match the input resolution
19 up_out = F.interpolate(out, size=[x.size(2), x.size(3)], mode='bilinear', align_corners=False)
20
21 # L2-normalize the first three channels / ensure positive value for concentration parameters (kappa)
22 up_out = norm_normalize(up_out)
23 return up_out
24
25 def get_1x_lr_params(self): # lr/10 learning rate
26 return self.encoder.parameters()

Callers

nothing calls this directly

Calls 1

norm_normalizeFunction · 0.90

Tested by

no test coverage detected