modules/inpaint/lama.py

import torch
import torch.nn as nn
from torch import Tensor
import numpy as np
import cv2

from .ffc import FFC_BN_ACT

def get_activation(kind='tanh'):
    if kind == 'tanh':
        return nn.Tanh()
    if kind == 'sigmoid':
        return nn.Sigmoid()
    if kind is False:
        return nn.Identity()
    raise ValueError(f'Unknown activation kind {kind}')

class FFCResnetBlock(nn.Module):
    def __init__(self, dim, padding_type, norm_layer, activation_layer=nn.ReLU, dilation=1,
                 inline=False, **conv_kwargs):
        super().__init__()
        self.conv1 = FFC_BN_ACT(dim, dim, kernel_size=3, padding=dilation, dilation=dilation,
                                norm_layer=norm_layer,
                                activation_layer=activation_layer,
                                padding_type=padding_type,
                                **conv_kwargs)
        self.conv2 = FFC_BN_ACT(dim, dim, kernel_size=3, padding=dilation, dilation=dilation,
                                norm_layer=norm_layer,
                                activation_layer=activation_layer,
                                padding_type=padding_type,
                                **conv_kwargs)
        self.inline = inline

    def forward(self, x):
        if self.inline:
            x_l, x_g = x[:, :-self.conv1.ffc.global_in_num], x[:, -self.conv1.ffc.global_in_num:]
        else:
            x_l, x_g = x if type(x) is tuple else (x, 0)

        id_l, id_g = x_l, x_g

        x_l, x_g = self.conv1((x_l, x_g))
        x_l, x_g = self.conv2((x_l, x_g))

        x_l, x_g = id_l + x_l, id_g + x_g
        out = x_l, x_g
        if self.inline:
            out = torch.cat(out, dim=1)
        return out


class ConcatTupleLayer(nn.Module):
    def forward(self, x):
        assert isinstance(x, tuple)
        x_l, x_g = x
        assert torch.is_tensor(x_l) or torch.is_tensor(x_g)
        if not torch.is_tensor(x_g):
            return x_l
        return torch.cat(x, dim=1)


class FFCResNetGenerator(nn.Module):
    def __init__(self, input_nc=4, output_nc=3, ngf=64, n_downsampling=3, n_blocks=9, norm_layer=nn.BatchNorm2d,
                 padding_type='reflect', activation_layer=nn.ReLU,
                 up_norm_layer=nn.BatchNorm2d, up_activation=nn.ReLU(True),
                 init_conv_kwargs={}, downsample_conv_kwargs={}, resnet_conv_kwargs={}, spatial_transform_kwargs={},
                 add_out_act=True, max_features=1024, out_ffc=False, out_ffc_kwargs={}):
        assert (n_blocks >= 0)
        super().__init__()

        model = [nn.ReflectionPad2d(3),
                 FFC_BN_ACT(input_nc, ngf, kernel_size=7, padding=0, norm_layer=norm_layer,
                            activation_layer=activation_layer, **init_conv_kwargs)]

        ### downsample
        for i in range(n_downsampling):
            mult = 2 ** i
            if i == n_downsampling - 1:
                cur_conv_kwargs = dict(downsample_conv_kwargs)
                cur_conv_kwargs['ratio_gout'] = resnet_conv_kwargs.get('ratio_gin', 0)
            else:
                cur_conv_kwargs = downsample_conv_kwargs
            model += [FFC_BN_ACT(min(max_features, ngf * mult),
                                 min(max_features, ngf * mult * 2),
                                 kernel_size=3, stride=2, padding=1,
                                 norm_layer=norm_layer,
                                 activation_layer=activation_layer,
                                 **cur_conv_kwargs)]

        mult = 2 ** n_downsampling
        feats_num_bottleneck = min(max_features, ngf * mult)

        ### resnet blocks
        for i in range(n_blocks):
            cur_resblock = FFCResnetBlock(feats_num_bottleneck, padding_type=padding_type, activation_layer=activation_layer,
                                          norm_layer=norm_layer, **resnet_conv_kwargs)
            model += [cur_resblock]

        model += [ConcatTupleLayer()]

        ### upsample
        for i in range(n_downsampling):
            mult = 2 ** (n_downsampling - i)
            model += [nn.ConvTranspose2d(min(max_features, ngf * mult),
                                         min(max_features, int(ngf * mult / 2)),
                                         kernel_size=3, stride=2, padding=1, output_padding=1),
                      up_norm_layer(min(max_features, int(ngf * mult / 2))),
                      up_activation]

        if out_ffc:
            model += [FFCResnetBlock(ngf, padding_type=padding_type, activation_layer=activation_layer,
                                     norm_layer=norm_layer, inline=True, **out_ffc_kwargs)]

        model += [nn.ReflectionPad2d(3),
                  nn.Conv2d(ngf, output_nc, kernel_size=7, padding=0)]
        if add_out_act:
            model.append(get_activation('tanh' if add_out_act is True else add_out_act))
        self.model = nn.Sequential(*model)

    def forward(self, img, mask, rel_pos=None, direct=None) -> Tensor:
        masked_img = torch.cat([img * (1 - mask), mask], dim=1)
        if rel_pos is None:
            return self.model(masked_img)
        else:
            
            x_l, x_g = self.model[:2](masked_img)
            x_l = x_l.to(torch.float32)
            x_l += rel_pos
            x_l += direct
            return self.model[2:]((x_l, x_g))


class NLayerDiscriminator(nn.Module):
    def __init__(self, input_nc=3, ndf=64, n_layers=4, norm_layer=nn.BatchNorm2d,):
        super().__init__()
        self.n_layers = n_layers

        kw = 4
        padw = int(np.ceil((kw-1.0)/2))
        sequence = [[nn.Conv2d(input_nc, ndf, kernel_size=kw, stride=2, padding=padw),
                     nn.LeakyReLU(0.2, True)]]

        nf = ndf
        for n in range(1, n_layers):
            nf_prev = nf
            nf = min(nf * 2, 512)

            cur_model = []
            cur_model += [
                nn.Conv2d(nf_prev, nf, kernel_size=kw, stride=2, padding=padw),
                norm_layer(nf),
                nn.LeakyReLU(0.2, True)
            ]
            sequence.append(cur_model)

        nf_prev = nf
        nf = min(nf * 2, 512)

        cur_model = []
        cur_model += [
            nn.Conv2d(nf_prev, nf, kernel_size=kw, stride=1, padding=padw),
            norm_layer(nf),
            nn.LeakyReLU(0.2, True)
        ]
        sequence.append(cur_model)

        sequence += [[nn.Conv2d(nf, 1, kernel_size=kw, stride=1, padding=padw)]]

        for n in range(len(sequence)):
            setattr(self, 'model'+str(n), nn.Sequential(*sequence[n]))

    def get_all_activations(self, x):
        res = [x]
        for n in range(self.n_layers + 2):
            model = getattr(self, 'model' + str(n))
            res.append(model(res[-1]))
        return res[1:]

    def forward(self, x):
        act = self.get_all_activations(x)
        return act[-1], act[:-1]

def set_requires_grad(module, value):
    for param in module.parameters():
        param.requires_grad = value


class MaskedSinusoidalPositionalEmbedding(nn.Embedding):
    """This module produces sinusoidal positional embeddings of any length."""

    def __init__(self, num_embeddings: int, embedding_dim: int):
        super().__init__(num_embeddings, embedding_dim)
        self.weight = self._init_weight(self.weight)

    @staticmethod
    def _init_weight(out: nn.Parameter):
        """
        Identical to the XLM create_sinusoidal_embeddings except features are not interleaved. The cos features are in
        the 2nd half of the vector. [dim // 2:]
        """
        n_pos, dim = out.shape
        position_enc = np.array(
            [[pos / np.power(10000, 2 * (j // 2) / dim) for j in range(dim)] for pos in range(n_pos)]
        )
        out.requires_grad = False  # set early to avoid an error in pytorch-1.8+
        sentinel = dim // 2 if dim % 2 == 0 else (dim // 2) + 1
        out[:, 0:sentinel] = torch.FloatTensor(np.sin(position_enc[:, 0::2]))
        out[:, sentinel:] = torch.FloatTensor(np.cos(position_enc[:, 1::2]))
        out.detach_()
        return out

    @torch.no_grad()
    def forward(self, input_ids):
        """`input_ids` is expected to be [bsz x seqlen]."""
        return super().forward(input_ids)


class MultiLabelEmbedding(nn.Module):
    def __init__(self, num_positions: int, embedding_dim: int):
        super().__init__()
        self.weight = nn.Parameter(torch.Tensor(num_positions, embedding_dim))
        self.reset_parameters()

    def reset_parameters(self):
        nn.init.normal_(self.weight)

    def forward(self, input_ids):
        # input_ids:[B,HW,4](onehot)
        out = torch.matmul(input_ids, self.weight)  # [B,HW,dim]
        return out


class MPE(nn.Module):
    def __init__(self):
        super().__init__()
        self.rel_pos_emb = MaskedSinusoidalPositionalEmbedding(num_embeddings=128,
                                                                embedding_dim=64)
        self.direct_emb = MultiLabelEmbedding(num_positions=4, embedding_dim=64)
        self.alpha5 = nn.Parameter(torch.tensor(0, dtype=torch.float32), requires_grad=True)
        self.alpha6 = nn.Parameter(torch.tensor(0, dtype=torch.float32), requires_grad=True)

    def forward(self, rel_pos=None, direct=None):
        b, h, w = rel_pos.shape
        rel_pos = rel_pos.reshape(b, h * w)
        rel_pos_emb = self.rel_pos_emb(rel_pos).reshape(b, h, w, -1).permute(0, 3, 1, 2) * self.alpha5
        direct = direct.reshape(b, h * w, 4).to(torch.float32)
        direct_emb = self.direct_emb(direct).reshape(b, h, w, -1).permute(0, 3, 1, 2) * self.alpha6
        return rel_pos_emb, direct_emb


class LamaFourier:
    def __init__(self, build_discriminator=True, use_mpe=False, large_arch: bool = False) -> None:
        # super().__init__()

        n_blocks = 9
        if large_arch:
            n_blocks = 18
        
        self.generator = FFCResNetGenerator(4, 3, add_out_act='sigmoid', 
                            n_blocks = n_blocks,
                            init_conv_kwargs={
                            'ratio_gin': 0,
                            'ratio_gout': 0,
                            'enable_lfu': False
                        }, downsample_conv_kwargs={
                            'ratio_gin': 0,
                            'ratio_gout': 0,
                            'enable_lfu': False
                        }, resnet_conv_kwargs={
                            'ratio_gin': 0.75,
                            'ratio_gout': 0.75,
                            'enable_lfu': False
                        }, 
                    )
        
        self.discriminator = NLayerDiscriminator() if build_discriminator else None
        self.inpaint_only = False
        if use_mpe:
            self.mpe = MPE()
        else:
            self.mpe = None

    def train_generator(self):
        self.inpaint_only = False
        self.forward_generator = True
        self.forward_discriminator = False
        self.generator.train()
        self.discriminator.eval()
        set_requires_grad(self.discriminator, False)
        set_requires_grad(self.generator, True)
        if self.mpe is not None:
            set_requires_grad(self.mpe, True)

    def train_discriminator(self):
        self.inpaint_only = False
        self.forward_generator = False
        self.forward_discriminator = True
        self.discriminator.train()
        self.generator.eval()
        set_requires_grad(self.discriminator, True)
        set_requires_grad(self.generator, False)
        if self.mpe is not None:
            set_requires_grad(self.mpe, False)

    def to(self, device):
        self.generator.to(device)
        if self.discriminator is not None:
            self.discriminator.to(device)
        if self.mpe is not None:
            self.mpe.to(device)

    def eval(self):
        self.inpaint_only = True
        self.generator.eval()
        if self.mpe is not None:
            self.mpe.eval()
        return self
        

    def __call__(self, img: Tensor, mask: Tensor, rel_pos=None, direct=None):

        if self.mpe is not None:
            rel_pos, direct = self.mpe(rel_pos, direct)
        else:
            rel_pos, direct = None, None
        predicted_img = self.generator(img, mask, rel_pos, direct)

        if self.inpaint_only:
            return predicted_img * mask + (1 - mask) * img

        if self.forward_discriminator:
            predicted_img = predicted_img.detach()
            img.requires_grad = True


        discr_real_pred, discr_real_features = self.discriminator(img)
        discr_fake_pred, discr_fake_features = self.discriminator(predicted_img)
        # fp = discr_fake_pred.detach().mean()
    
        if self.forward_discriminator:
            return  {
                'predicted_img': predicted_img, 
                'discr_real_pred': discr_real_pred, 
                'discr_fake_pred':discr_fake_pred
            }
        else:
            return  {
                'predicted_img': predicted_img, 
                'discr_real_features': discr_real_features, 
                'discr_fake_features': discr_fake_features, 
                'discr_fake_pred': discr_fake_pred
            }

    def load_masked_position_encoding(self, mask):
        mask = (mask * 255).astype(np.uint8)
        ones_filter = np.ones((3, 3), dtype=np.float32)
        d_filter1 = np.array([[1, 1, 0], [1, 1, 0], [0, 0, 0]], dtype=np.float32)
        d_filter2 = np.array([[0, 0, 0], [1, 1, 0], [1, 1, 0]], dtype=np.float32)
        d_filter3 = np.array([[0, 1, 1], [0, 1, 1], [0, 0, 0]], dtype=np.float32)
        d_filter4 = np.array([[0, 0, 0], [0, 1, 1], [0, 1, 1]], dtype=np.float32)
        str_size = 256
        pos_num = 128

        ori_mask = mask.copy()
        ori_h, ori_w = ori_mask.shape[0:2]
        ori_mask = ori_mask / 255
        mask = cv2.resize(mask, (str_size, str_size), interpolation=cv2.INTER_AREA)
        mask[mask > 0] = 255
        h, w = mask.shape[0:2]
        mask3 = mask.copy()
        mask3 = 1. - (mask3 / 255.0)
        pos = np.zeros((h, w), dtype=np.int32)
        direct = np.zeros((h, w, 4), dtype=np.int32)
        i = 0

        if mask3.max() > 0:
            # otherwise it will cause infinity loop
        
            while np.sum(1 - mask3) > 0:
                i += 1
                mask3_ = cv2.filter2D(mask3, -1, ones_filter)
                mask3_[mask3_ > 0] = 1
                sub_mask = mask3_ - mask3
                pos[sub_mask == 1] = i

                m = cv2.filter2D(mask3, -1, d_filter1)
                m[m > 0] = 1
                m = m - mask3
                direct[m == 1, 0] = 1

                m = cv2.filter2D(mask3, -1, d_filter2)
                m[m > 0] = 1
                m = m - mask3
                direct[m == 1, 1] = 1

                m = cv2.filter2D(mask3, -1, d_filter3)
                m[m > 0] = 1
                m = m - mask3
                direct[m == 1, 2] = 1

                m = cv2.filter2D(mask3, -1, d_filter4)
                m[m > 0] = 1
                m = m - mask3
                direct[m == 1, 3] = 1

                mask3 = mask3_

        abs_pos = pos.copy()
        rel_pos = pos / (str_size / 2)  # to 0~1 maybe larger than 1
        rel_pos = (rel_pos * pos_num).astype(np.int32)
        rel_pos = np.clip(rel_pos, 0, pos_num - 1)

        if ori_w != w or ori_h != h:
            rel_pos = cv2.resize(rel_pos, (ori_w, ori_h), interpolation=cv2.INTER_NEAREST)
            rel_pos[ori_mask == 0] = 0
            direct = cv2.resize(direct, (ori_w, ori_h), interpolation=cv2.INTER_NEAREST)
            direct[ori_mask == 0, :] = 0

        return rel_pos, abs_pos, direct

def load_lama_mpe(model_path, device, use_mpe=True, large_arch: bool = False) -> LamaFourier:
    model = LamaFourier(build_discriminator=False, use_mpe=use_mpe, large_arch=large_arch)
    sd = torch.load(model_path, map_location = 'cpu')
    model.generator.load_state_dict(sd['gen_state_dict'])
    if use_mpe:
        model.mpe.load_state_dict(sd['str_state_dict'])
    model.eval().to(device)
    return model