MCPcopy Create free account
hub / github.com/VCIP-RGBD/DFormer / encode_decode

Method encode_decode

models/builder.py:225–239  ·  view source on GitHub ↗

Encode images with backbone and decode into a semantic segmentation map of the same size as input.

(self, rgb, modal_x)

Source from the content-addressed store, hash-verified

223 )
224
225 def encode_decode(self, rgb, modal_x):
226 """Encode images with backbone and decode into a semantic segmentation
227 map of the same size as input."""
228 orisize = rgb.shape
229 # print('builder',rgb.shape,modal_x.shape)
230 x = self.backbone(rgb, modal_x)
231 if len(x) == 2: # if output is (rgb,depth) only use rgb
232 x = x[0]
233 out = self.decode_head.forward(x)
234 out = F.interpolate(out, size=orisize[-2:], mode="bilinear", align_corners=False)
235 if self.aux_head:
236 aux_fm = self.aux_head(x[0][self.aux_index])
237 aux_fm = F.interpolate(aux_fm, size=orisize[2:], mode="bilinear", align_corners=False)
238 return out, aux_fm
239 return out
240
241 def forward(self, rgb, modal_x=None, label=None):
242 # print('builder',rgb.shape,modal_x.shape)

Callers 1

forwardMethod · 0.95

Calls 1

forwardMethod · 0.45

Tested by

no test coverage detected