def forward(self, c3, c4, c5):
b = c3.size(0)
p51 = self.conv51(c5)
p52 = self.conv52(p51)
p5 = self.pred5(p52)
p41 = upsample_concat(self.lat5(p51), c4)
p42 = self.conv41(p41)
p43 = self.conv42(p42)
p4 = self.pred4(p43)
p31 = upsample_concat(self.lat4(p42), c3)
p32 = self.conv31(p31)
p33 = self.conv32(p32)
p3 = self.pred3(p33)
preds = [p3, p4, p5]
loc_preds = []
obj_preds = []
cls_preds = []
log_var_preds = []
for p in preds:
p = p.permute(0, 3, 2,
1).contiguous().view(b, -1, 9 + self.num_classes)
loc_preds.append(p[..., :4])
log_var_preds.append(p[..., 4:8])
obj_preds.append(p[..., 8])
cls_preds.append(p[..., 9:])
loc_p = torch.cat(loc_preds, dim=1)
obj_p = torch.cat(obj_preds, dim=1)
cls_p = torch.cat(cls_preds, dim=1)
log_var_p = torch.cat(log_var_preds, dim=1)
return loc_p, obj_p, cls_p, log_var_p
def forward(self, c, p):
if self.aggregate == 'cat':
p = upsample_concat(p, self.lat(c))
else:
p = upsample_add(p, self.lat(c))
p = self.conv(p)
return p
def forward(self, *cs):
ps = []
p1 = self.convs[0](cs[-1])
p2 = self.outs[0](p1)
ps.append(p2)
for lat, conv, out, c in zip(self.lats, self.convs[1:], self.outs[1:],
reversed(cs[:-1])):
c = upsample_concat(lat(p1), c)
p1 = conv(c)
p2 = out(p1)
ps.append(p2)
return tuple(reversed(ps))
def forward(self, x1, x2):
x1 = self.conv1(x1)
x2 = self.conv2(x2)
return upsample_concat(x2, x1)
