|
|
import torch |
|
|
from torch.nn import functional as F |
|
|
import math |
|
|
|
|
|
|
|
|
def upfirdn2d_native(input, kernel, up_x, up_y, down_x, down_y, pad_x0, pad_x1, pad_y0, pad_y1): |
|
|
_, minor, in_h, in_w = input.shape |
|
|
kernel_h, kernel_w = kernel.shape |
|
|
|
|
|
out = input.view(-1, minor, in_h, 1, in_w, 1) |
|
|
out = F.pad(out, [0, up_x - 1, 0, 0, 0, up_y - 1, 0, 0]) |
|
|
out = out.view(-1, minor, in_h * up_y, in_w * up_x) |
|
|
|
|
|
out = F.pad(out, [max(pad_x0, 0), max(pad_x1, 0), max(pad_y0, 0), max(pad_y1, 0)]) |
|
|
out = out[:, :, max(-pad_y0, 0): out.shape[2] - max(-pad_y1, 0), max(-pad_x0, 0): out.shape[3] - max(-pad_x1, 0)] |
|
|
|
|
|
out = out.reshape([-1, 1, in_h * up_y + pad_y0 + pad_y1, in_w * up_x + pad_x0 + pad_x1]) |
|
|
w = torch.flip(kernel, [0, 1]).view(1, 1, kernel_h, kernel_w) |
|
|
out = F.conv2d(out, w) |
|
|
out = out.reshape(-1, minor, in_h * up_y + pad_y0 + pad_y1 - kernel_h + 1, in_w * up_x + pad_x0 + pad_x1 - kernel_w + 1) |
|
|
return out[:, :, ::down_y, ::down_x] |
|
|
|
|
|
def upfirdn2d(input, kernel, up=1, down=1, pad=(0, 0)): |
|
|
return upfirdn2d_native(input, kernel, up, up, down, down, pad[0], pad[1], pad[0], pad[1]) |
|
|
|
|
|
|
|
|
class FusedLeakyReLU(torch.nn.Module): |
|
|
def __init__(self, channel, negative_slope=0.2, scale=2 ** 0.5): |
|
|
super().__init__() |
|
|
self.bias = torch.nn.Parameter(torch.zeros(1, channel, 1, 1)) |
|
|
self.negative_slope = negative_slope |
|
|
self.scale = scale |
|
|
|
|
|
def forward(self, input): |
|
|
return fused_leaky_relu(input, self.bias, self.negative_slope, self.scale) |
|
|
|
|
|
def fused_leaky_relu(input, bias, negative_slope=0.2, scale=2 ** 0.5): |
|
|
return F.leaky_relu(input + bias, negative_slope) * scale |
|
|
|
|
|
class Blur(torch.nn.Module): |
|
|
def __init__(self, kernel, pad): |
|
|
super().__init__() |
|
|
kernel = torch.tensor(kernel, dtype=torch.float32) |
|
|
kernel = kernel[None, :] * kernel[:, None] |
|
|
kernel = kernel / kernel.sum() |
|
|
self.register_buffer('kernel', kernel) |
|
|
self.pad = pad |
|
|
|
|
|
def forward(self, input): |
|
|
return upfirdn2d(input, self.kernel, pad=self.pad) |
|
|
|
|
|
|
|
|
class ScaledLeakyReLU(torch.nn.Module): |
|
|
def __init__(self, negative_slope=0.2): |
|
|
super().__init__() |
|
|
self.negative_slope = negative_slope |
|
|
|
|
|
def forward(self, input): |
|
|
return F.leaky_relu(input, negative_slope=self.negative_slope) |
|
|
|
|
|
|
|
|
class EqualConv2d(torch.nn.Module): |
|
|
def __init__(self, in_channel, out_channel, kernel_size, stride=1, padding=0, bias=True): |
|
|
super().__init__() |
|
|
self.weight = torch.nn.Parameter(torch.randn(out_channel, in_channel, kernel_size, kernel_size)) |
|
|
self.scale = 1 / math.sqrt(in_channel * kernel_size ** 2) |
|
|
self.stride = stride |
|
|
self.padding = padding |
|
|
self.bias = torch.nn.Parameter(torch.zeros(out_channel)) if bias else None |
|
|
|
|
|
def forward(self, input): |
|
|
return F.conv2d(input, self.weight * self.scale, bias=self.bias, stride=self.stride, padding=self.padding) |
|
|
|
|
|
|
|
|
class EqualLinear(torch.nn.Module): |
|
|
def __init__(self, in_dim, out_dim, bias=True, bias_init=0, lr_mul=1, activation=None): |
|
|
super().__init__() |
|
|
self.weight = torch.nn.Parameter(torch.randn(out_dim, in_dim).div_(lr_mul)) |
|
|
self.bias = torch.nn.Parameter(torch.zeros(out_dim).fill_(bias_init)) if bias else None |
|
|
self.activation = activation |
|
|
self.scale = (1 / math.sqrt(in_dim)) * lr_mul |
|
|
self.lr_mul = lr_mul |
|
|
|
|
|
def forward(self, input): |
|
|
if self.activation: |
|
|
out = F.linear(input, self.weight * self.scale) |
|
|
return fused_leaky_relu(out, self.bias * self.lr_mul) |
|
|
return F.linear(input, self.weight * self.scale, bias=self.bias * self.lr_mul) |
|
|
|
|
|
|
|
|
class ConvLayer(torch.nn.Sequential): |
|
|
def __init__(self, in_channel, out_channel, kernel_size, downsample=False, blur_kernel=[1, 3, 3, 1], bias=True, activate=True): |
|
|
layers = [] |
|
|
|
|
|
if downsample: |
|
|
factor = 2 |
|
|
p = (len(blur_kernel) - factor) + (kernel_size - 1) |
|
|
layers.append(Blur(blur_kernel, pad=((p + 1) // 2, p // 2))) |
|
|
stride, padding = 2, 0 |
|
|
else: |
|
|
stride, padding = 1, kernel_size // 2 |
|
|
|
|
|
layers.append(EqualConv2d(in_channel, out_channel, kernel_size, padding=padding, stride=stride, bias=bias and not activate)) |
|
|
|
|
|
if activate: |
|
|
layers.append(FusedLeakyReLU(out_channel) if bias else ScaledLeakyReLU(0.2)) |
|
|
|
|
|
super().__init__(*layers) |
|
|
|
|
|
|
|
|
class ResBlock(torch.nn.Module): |
|
|
def __init__(self, in_channel, out_channel): |
|
|
super().__init__() |
|
|
self.conv1 = ConvLayer(in_channel, in_channel, 3) |
|
|
self.conv2 = ConvLayer(in_channel, out_channel, 3, downsample=True) |
|
|
self.skip = ConvLayer(in_channel, out_channel, 1, downsample=True, activate=False, bias=False) |
|
|
|
|
|
def forward(self, input): |
|
|
out = self.conv2(self.conv1(input)) |
|
|
skip = self.skip(input) |
|
|
return (out + skip) / math.sqrt(2) |
|
|
|
|
|
|
|
|
class AppearanceEncoder(torch.nn.Module): |
|
|
def __init__(self, w_dim=512): |
|
|
super().__init__() |
|
|
|
|
|
self.convs = torch.nn.ModuleList([ |
|
|
ConvLayer(3, 32, 1), ResBlock(32, 64), |
|
|
ResBlock(64, 128), ResBlock(128, 256), |
|
|
ResBlock(256, 512), ResBlock(512, 512), |
|
|
ResBlock(512, 512), ResBlock(512, 512), |
|
|
EqualConv2d(512, w_dim, 4, padding=0, bias=False) |
|
|
]) |
|
|
|
|
|
def forward(self, x): |
|
|
for conv in self.convs: |
|
|
x = conv(x) |
|
|
return x.squeeze((-2, -1)) |
|
|
|
|
|
class MotionEncoder(torch.nn.Module): |
|
|
def __init__(self, dim=512, motion_dim=20): |
|
|
super().__init__() |
|
|
self.net_app = AppearanceEncoder(dim) |
|
|
self.fc = torch.nn.Sequential(*[EqualLinear(dim, dim) for _ in range(4)] + [EqualLinear(dim, motion_dim)]) |
|
|
|
|
|
def encode_motion(self, x): |
|
|
return self.fc(self.net_app(x)) |
|
|
|
|
|
class MotionProjector(torch.nn.Module): |
|
|
def __init__(self, m_dim): |
|
|
super().__init__() |
|
|
self.weight = torch.nn.Parameter(torch.randn(512, m_dim)) |
|
|
self.motion_dim = m_dim |
|
|
|
|
|
def forward(self, input): |
|
|
stabilized_weight = self.weight + 1e-8 * torch.eye(512, self.motion_dim, device=self.weight.device, dtype=self.weight.dtype) |
|
|
Q, _ = torch.linalg.qr(stabilized_weight) |
|
|
if input is None: |
|
|
return Q |
|
|
return torch.sum(input.unsqueeze(-1) * Q.T, dim=1) |
|
|
|
|
|
class MotionDecoder(torch.nn.Module): |
|
|
def __init__(self, m_dim): |
|
|
super().__init__() |
|
|
self.direction = MotionProjector(m_dim) |
|
|
|
|
|
class MotionExtractor(torch.nn.Module): |
|
|
def __init__(self, s_dim=512, m_dim=20): |
|
|
super().__init__() |
|
|
self.enc = MotionEncoder(s_dim, m_dim) |
|
|
self.dec = MotionDecoder(m_dim) |
|
|
|
|
|
def forward(self, img): |
|
|
motion_feat = self.enc.encode_motion(img) |
|
|
return self.dec.direction(motion_feat) |