liav
/
lut_reproduce
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix

Browse Source
main
vlpr 7 months ago
parent 96d27bfd05
commit 72ac9d99fd
1 changed files with 21 additions and 21 deletions
Show Stats Download Patch File Download Diff File

42
src/models/sdylut.py
Unescape Escape View File

@ -78,15 +78,15 @@ class SDYLutx2(nn.Module):
super(SDYLutx2, self).__init__() super(SDYLutx2, self).__init__()
self.scale = scale self.scale = scale
self.quantization_interval = quantization_interval self.quantization_interval = quantization_interval
self._extract_pattern_S = PercievePattern(receptive_field_idxes=[[0,0],[0,1],[1,0],[1,1]], center=[0,0], window_size=3) self._extract_pattern_S = PercievePattern(receptive_field_idxes=[[0,0],[0,1],[1,0],[1,1]], center=[0,0], window_size=2)
self._extract_pattern_D = PercievePattern(receptive_field_idxes=[[0,0],[2,0],[0,2],[2,2]], center=[0,0], window_size=3) self._extract_pattern_D = PercievePattern(receptive_field_idxes=[[0,0],[2,0],[0,2],[2,2]], center=[0,0], window_size=3)
self._extract_pattern_Y = PercievePattern(receptive_field_idxes=[[0,0],[1,1],[1,2],[2,1]], center=[0,0], window_size=3) self._extract_pattern_Y = PercievePattern(receptive_field_idxes=[[0,0],[1,1],[1,2],[2,1]], center=[0,0], window_size=3)
self.stageS_1 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage1_S = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (1,1)).type(torch.float32))
self.stageD_1 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage1_D = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (1,1)).type(torch.float32))
self.stageY_1 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage1_Y = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (1,1)).type(torch.float32))
self.stageS_2 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage2_S = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32))
self.stageD_2 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage2_D = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32))
self.stageY_2 = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32)) self.stage2_Y = nn.Parameter(torch.randint(0, 255, size=(256//quantization_interval+1,)*4 + (scale,scale)).type(torch.float32))
@staticmethod @staticmethod
def init_from_lut( def init_from_lut(
@ -106,28 +106,27 @@ class SDYLutx2(nn.Module):
def forward(self, x): def forward(self, x):
b,c,h,w = x.shape b,c,h,w = x.shape
x = x.view(b*c, 1, h, w).type(torch.float32) x = x.view(b*c, 1, h, w).type(torch.float32)
output = torch.zeros([b*c, 1, h, w], dtype=x.dtype, device=x.device) output = torch.zeros([b*c, 1, h, w], dtype=x.dtype, device=x.device)
for rotations_count in range(4): for rotations_count in range(4):
rotated = torch.rot90(x, k=rotations_count, dims=[-2, -1]) rotated = torch.rot90(x, k=rotations_count, dims=[-2, -1])
rb,rc,rh,rw = rotated.shape rb,rc,rh,rw = rotated.shape
s = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_S(rotated), lut=self.stage1_S) s = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_S(rotated), lut=self.stage1_S)
s = s.view(rb*rc, 1, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh, rw) s = s.view(rb, rc, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb, rc, rh, rw)
s = torch.rot90(s, k=-rotations_count, dims=[-2, -1]) s = torch.rot90(s, k=-rotations_count, dims=[-2, -1])
output += s output += s
d = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_D(rotated), lut=self.stage1_D) d = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_D(rotated), lut=self.stage1_D)
d = d.view(rb*rc, 1, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh, rw) d = d.view(rb, rc, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb, rc, rh, rw)
d = torch.rot90(d, k=-rotations_count, dims=[-2, -1]) d = torch.rot90(d, k=-rotations_count, dims=[-2, -1])
output += d output += d
y = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_Y(rotated), lut=self.stage1_Y) y = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_Y(rotated), lut=self.stage1_Y)
y = y.view(rb*rc, 1, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh, rw) y = y.view(rb, rc, rh, rw, 1, 1).permute(0,1,2,4,3,5).reshape(rb, rc, rh, rw)
y = torch.rot90(y, k=-rotations_count, dims=[-2, -1]) y = torch.rot90(y, k=-rotations_count, dims=[-2, -1])
output += y output += y
output /= 4*3 output /= 4*3
output = output.view(b, c, h, w)
x = output x = output
output = torch.zeros([b*c, 1, h*self.scale, w*self.scale], dtype=x.dtype, device=x.device) output = torch.zeros([b*c, 1, h*self.scale, w*self.scale], dtype=x.dtype, device=x.device)
@ -136,32 +135,33 @@ class SDYLutx2(nn.Module):
rb,rc,rh,rw = rotated.shape rb,rc,rh,rw = rotated.shape
s = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_S(rotated), lut=self.stage2_S) s = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_S(rotated), lut=self.stage2_S)
s = s.view(rb*rc, 1, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh*self.scale, rw*self.scale) s = s.view(rb, rc, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb, rc, rh*self.scale, rw*self.scale)
s = torch.rot90(s, k=-rotations_count, dims=[-2, -1]) s = torch.rot90(s, k=-rotations_count, dims=[-2, -1])
output += s output += s
d = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_D(rotated), lut=self.stage2_D) d = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_D(rotated), lut=self.stage2_D)
d = d.view(rb*rc, 1, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh*self.scale, rw*self.scale) d = d.view(rb, rc, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb, rc, rh*self.scale, rw*self.scale)
d = torch.rot90(d, k=-rotations_count, dims=[-2, -1]) d = torch.rot90(d, k=-rotations_count, dims=[-2, -1])
output += d output += d
y = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_Y(rotated), lut=self.stage2_Y) y = forward_unfolded_2x2_input_SxS_output(index=self._extract_pattern_Y(rotated), lut=self.stage2_Y)
y = y.view(rb*rc, 1, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb*rc, 1, rh*self.scale, rw*self.scale) y = y.view(rb, rc, rh, rw, self.scale, self.scale).permute(0,1,2,4,3,5).reshape(rb, rc, rh*self.scale, rw*self.scale)
y = torch.rot90(y, k=-rotations_count, dims=[-2, -1]) y = torch.rot90(y, k=-rotations_count, dims=[-2, -1])
output += y output += y
output /= 4*3 output /= 4*3
output = output.view(b, c, h*self.scale, w*self.scale) output = output.view(b, c, h*self.scale, w*self.scale)
return output return output
def __repr__(self): def __repr__(self):
return f"{self.__class__.__name__}" + \ return f"{self.__class__.__name__}" + \
f"\n stageS_1 size: {self.stageS_1.shape}" + \ f"\n stage1_S size: {self.stage1_S.shape}" + \
f"\n stageD_1 size: {self.stageD_1.shape}" + \ f"\n stage1_D size: {self.stage1_D.shape}" + \
f"\n stageY_1 size: {self.stageY_1.shape}" + \ f"\n stage1_Y size: {self.stage1_Y.shape}" + \
f"\n stageS_2 size: {self.stageS_2.shape}" + \ f"\n stage2_S size: {self.stage2_S.shape}" + \
f"\n stageD_2 size: {self.stageD_2.shape}" + \ f"\n stage2_D size: {self.stage2_D.shape}" + \
f"\n stageY_2 size: {self.stageY_2.shape}" f"\n stage2_Y size: {self.stage2_Y.shape}"

Write Preview
Loading…
Cancel
Save