cisen/model/layers.py

import math

import torch
import torch.nn as nn
import torch.nn.functional as F
# import open_clip

def conv_layer(in_dim, out_dim, kernel_size=1, padding=0, stride=1):
    return nn.Sequential(
        nn.Conv2d(in_dim, out_dim, kernel_size, stride, padding, bias=False),
        nn.BatchNorm2d(out_dim), nn.ReLU(True))
    # return nn.Sequential(
    #     nn.Conv2d(in_dim, out_dim, kernel_size, stride, padding, bias=False),
    #     nn.LayerNorm(out_dim), nn.ReLU(True))


# def conv_layer_1(in_dim, out_dim, kernel_size=1, padding=0, stride=1):
#     return nn.Sequential(
#         nn.Conv2d(in_dim, out_dim, kernel_size, stride, padding, bias=False),
#         nn.LayerNorm(out_dim), nn.ReLU(True))

def linear_layer(in_dim, out_dim,bias=False):
    return nn.Sequential(nn.Linear(in_dim, out_dim, bias),
                         nn.BatchNorm1d(out_dim), nn.ReLU(True))
    # return nn.Sequential(nn.Linear(in_dim, out_dim, bias),
    #                      nn.LayerNorm(out_dim), nn.ReLU(True))
class AttentionPool2d(nn.Module):
    def __init__(self, spacial_dim: int, embed_dim: int, num_heads: int, output_dim: int = None):
        super().__init__()
        self.positional_embedding = nn.Parameter(torch.randn(spacial_dim ** 2 + 1, embed_dim) / embed_dim ** 0.5)
        self.k_proj = nn.Linear(embed_dim, embed_dim)
        self.q_proj = nn.Linear(embed_dim, embed_dim)
        self.v_proj = nn.Linear(embed_dim, embed_dim)
        self.c_proj = nn.Linear(embed_dim, output_dim or embed_dim)
        self.num_heads = num_heads

    def forward(self, x):
        x = x.flatten(start_dim=2).permute(2, 0, 1)  # NCHW -> (HW)NC
        x = torch.cat([x.mean(dim=0, keepdim=True), x], dim=0)  # (HW+1)NC
        x = x + self.positional_embedding[:, None, :].to(x.dtype)  # (HW+1)NC
        x, _ = F.multi_head_attention_forward(
            query=x[:1], key=x, value=x,
            embed_dim_to_check=x.shape[-1],
            num_heads=self.num_heads,
            q_proj_weight=self.q_proj.weight,
            k_proj_weight=self.k_proj.weight,
            v_proj_weight=self.v_proj.weight,
            in_proj_weight=None,
            in_proj_bias=torch.cat([self.q_proj.bias, self.k_proj.bias, self.v_proj.bias]),
            bias_k=None,
            bias_v=None,
            add_zero_attn=False,
            dropout_p=0,
            out_proj_weight=self.c_proj.weight,
            out_proj_bias=self.c_proj.bias,
            use_separate_proj_weight=True,
            training=self.training,
            need_weights=False
        )
        return x.squeeze(0)

# class AttentionPool2d(nn.Module):
#     def __init__(self, spacial_dim: int, embed_dim: int, num_heads: int, output_dim: int = None):
#         super().__init__()
#         self.positional_embedding = nn.Parameter(torch.randn(spacial_dim ** 2 + 1, embed_dim) / embed_dim ** 0.5)
#         self.k_proj = nn.Linear(embed_dim, embed_dim)
#         self.q_proj = nn.Linear(embed_dim, embed_dim)
#         self.v_proj = nn.Linear(embed_dim, embed_dim)
#         self.c_proj = nn.Linear(embed_dim, output_dim or embed_dim)
#         self.num_heads = num_heads
#
#     def forward(self, x):
#         x = x.reshape(x.shape[0], x.shape[1], x.shape[2] * x.shape[3]).permute(2, 0, 1)  # NCHW -> (HW)NC
#         x = torch.cat([x.mean(dim=0, keepdim=True), x], dim=0)  # (HW+1)NC
#         x = x + self.positional_embedding[:, None, :].to(x.dtype)  # (HW+1)NC
#         x, _ = F.multi_head_attention_forward(
#             query=x, key=x, value=x,
#             embed_dim_to_check=x.shape[-1],
#             num_heads=self.num_heads,
#             q_proj_weight=self.q_proj.weight,
#             k_proj_weight=self.k_proj.weight,
#             v_proj_weight=self.v_proj.weight,
#             in_proj_weight=None,
#             in_proj_bias=torch.cat([self.q_proj.bias, self.k_proj.bias, self.v_proj.bias]),
#             bias_k=None,
#             bias_v=None,
#             add_zero_attn=False,
#             dropout_p=0,
#             out_proj_weight=self.c_proj.weight,
#             out_proj_bias=self.c_proj.bias,
#             use_separate_proj_weight=True,
#             training=self.training,
#             need_weights=False
#         )
#
#         return x[0]

class CoordConv(nn.Module):
    def __init__(self,
                 in_channels,
                 out_channels,
                 kernel_size=3,
                 padding=1,
                 stride=1):
        super().__init__()
        self.conv1 = conv_layer(in_channels + 2, out_channels, kernel_size,
                                padding, stride)

    def add_coord(self, input):
        b, _, h, w = input.size()
        x_range = torch.linspace(-1, 1, w, device=input.device)
        y_range = torch.linspace(-1, 1, h, device=input.device)
        y, x = torch.meshgrid(y_range, x_range)
        y = y.expand([b, 1, -1, -1])
        x = x.expand([b, 1, -1, -1])
        coord_feat = torch.cat([x, y], 1)
        input = torch.cat([input, coord_feat], 1)
        return input

    def forward(self, x):
        x = self.add_coord(x)
        x = self.conv1(x)
        return x

class TransformerDecoder(nn.Module):
    def __init__(self,
                 num_layers,
                 d_model,
                 nhead,
                 dim_ffn,
                 dropout,
                 return_intermediate=False):
        super().__init__()
        self.layers = nn.ModuleList([
            TransformerDecoderLayer(d_model=d_model,
                                    nhead=nhead,
                                    dim_feedforward=dim_ffn,
                                    dropout=dropout) for _ in range(num_layers)
        ])
        self.num_layers = num_layers
        self.norm = nn.LayerNorm(d_model)
        self.return_intermediate = return_intermediate

    @staticmethod
    def pos1d(d_model, length):
        """
        :param d_model: dimension of the model
        :param length: length of positions
        :return: length*d_model position matrix
        """
        if d_model % 2 != 0:
            raise ValueError("Cannot use sin/cos positional encoding with "
                             "odd dim (got dim={:d})".format(d_model))
        pe = torch.zeros(length, d_model)
        position = torch.arange(0, length).unsqueeze(1)
        div_term = torch.exp((torch.arange(0, d_model, 2, dtype=torch.float) *
                              -(math.log(10000.0) / d_model)))
        pe[:, 0::2] = torch.sin(position.float() * div_term)
        pe[:, 1::2] = torch.cos(position.float() * div_term)

        return pe.unsqueeze(1)  # n, 1, 512

    @staticmethod
    def pos2d(d_model, height, width):
        """
        :param d_model: dimension of the model
        :param height: height of the positions
        :param width: width of the positions
        :return: d_model*height*width position matrix
        """
        if d_model % 4 != 0:
            raise ValueError("Cannot use sin/cos positional encoding with "
                             "odd dimension (got dim={:d})".format(d_model))
        pe = torch.zeros(d_model, height, width)
        # Each dimension use half of d_model
        d_model = int(d_model / 2)
        div_term = torch.exp(
            torch.arange(0., d_model, 2) * -(math.log(10000.0) / d_model))
        pos_w = torch.arange(0., width).unsqueeze(1)
        pos_h = torch.arange(0., height).unsqueeze(1)
        pe[0:d_model:2, :, :] = torch.sin(pos_w * div_term).transpose(
            0, 1).unsqueeze(1).repeat(1, height, 1)
        pe[1:d_model:2, :, :] = torch.cos(pos_w * div_term).transpose(
            0, 1).unsqueeze(1).repeat(1, height, 1)
        pe[d_model::2, :, :] = torch.sin(pos_h * div_term).transpose(
            0, 1).unsqueeze(2).repeat(1, 1, width)
        pe[d_model + 1::2, :, :] = torch.cos(pos_h * div_term).transpose(
            0, 1).unsqueeze(2).repeat(1, 1, width)

        return pe.reshape(-1, 1, height * width).permute(2, 1, 0)  # hw, 1, 512

    def forward(self, vis, txt, pad_mask):
        '''
            vis: b, 512, h, w
            txt: b, L, 512
            pad_mask: b, L
        '''
        B, C, H, W = vis.size()
        _, L, D = txt.size()
        # position encoding
        vis_pos = self.pos2d(C, H, W)
        txt_pos = self.pos1d(D, L)
        # reshape & permute
        vis = vis.reshape(B, C, -1).permute(2, 0, 1)
        txt = txt.permute(1, 0, 2)
        # forward
        output = vis
        intermediate = []
        for layer in self.layers:
            output = layer(output, txt, vis_pos, txt_pos, pad_mask)
            if self.return_intermediate:
                # HW, b, 512 -> b, 512, HW
                intermediate.append(self.norm(output).permute(1, 2, 0))

        if self.norm is not None:
            # HW, b, 512 -> b, 512, HW
            output = self.norm(output).permute(1, 2, 0)
            if self.return_intermediate:
                intermediate.pop()
                intermediate.append(output)
                # [output1, output2, ..., output_n]
                return intermediate
            else:
                # b, 512, HW
                return output
        return output


class TransformerDecoderLayer(nn.Module):
    def __init__(self,
                 d_model=512,
                 nhead=9,
                 dim_feedforward=2048,
                 dropout=0.1):
        super().__init__()
        # Normalization Layer
        self.self_attn_norm = nn.LayerNorm(d_model)
        self.cross_attn_norm = nn.LayerNorm(d_model)
        # Attention Layer
        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)
        self.multihead_attn = nn.MultiheadAttention(d_model,
                                                    nhead,
                                                    dropout=dropout,
                                                    kdim=d_model,
                                                    vdim=d_model)
        # FFN
        self.ffn = nn.Sequential(nn.Linear(d_model, dim_feedforward),
                                 nn.ReLU(True), nn.Dropout(dropout),
                                 nn.LayerNorm(dim_feedforward),
                                 nn.Linear(dim_feedforward, d_model))
        # LayerNorm & Dropout
        self.norm1 = nn.LayerNorm(d_model)
        self.norm2 = nn.LayerNorm(d_model)
        self.norm3 = nn.LayerNorm(d_model)
        self.dropout1 = nn.Dropout(dropout)
        self.dropout2 = nn.Dropout(dropout)
        self.dropout3 = nn.Dropout(dropout)

    def with_pos_embed(self, tensor, pos):
        return tensor if pos is None else tensor + pos.to(tensor.device)

    def forward(self, vis, txt, vis_pos, txt_pos, pad_mask):
        '''
            vis: 26*26, b, 512
            txt: L, b, 512
            vis_pos: 26*26, 1, 512
            txt_pos: L, 1, 512
            pad_mask: b, L
        '''
        # Self-Attention
        vis2 = self.norm1(vis)
        q = k = self.with_pos_embed(vis2, vis_pos)
        vis2 = self.self_attn(q, k, value=vis2)[0]
        vis2 = self.self_attn_norm(vis2)
        vis = vis + self.dropout1(vis2)
        # Cross-Attention
        vis2 = self.norm2(vis)
        vis2 = self.multihead_attn(query=self.with_pos_embed(vis2, vis_pos),
                                   key=self.with_pos_embed(txt, txt_pos),
                                   value=txt,
                                   key_padding_mask=pad_mask)[0]
        vis2 = self.cross_attn_norm(vis2)
        vis = vis + self.dropout2(vis2)
        # FFN
        vis2 = self.norm3(vis)
        vis2 = self.ffn(vis2)
        vis = vis + self.dropout3(vis2)
        return vis

class Text_Projector(nn.Module):
    def __init__(self, args, in_channels=[512, 1024, 1024],
                 out_channels=[256, 512, 1024]):

        super(Text_Projector, self).__init__()

        self.proj = linear_layer(args, in_channels[2], out_channels[2])
        self.ReLU = nn.ReLU(True)

    def forward(self, text):

        text = self.ReLU(text + self.proj(text))

        return text

class Image_Projector(nn.Module):
    def __init__(self, args, in_channels=[512, 1024, 1024],
                 out_channels=[256, 512, 1024]):

        super(Image_Projector, self).__init__()

        self.proj = linear_layer(args, in_channels[0], out_channels[2])
        self.ReLU = nn.ReLU(True)

    def forward(self, image):

        image = self.ReLU(image + self.proj(image))

        return image

class Adapter(nn.Module):
    def __init__(self, c_in, reduction=4):
        super(Adapter, self).__init__()
        self.fc = nn.Sequential(
            nn.Linear(c_in, c_in // reduction, bias=False),
            nn.ReLU(inplace=True),
            nn.Linear(c_in // reduction, c_in, bias=False),
            nn.ReLU(inplace=True)
        )

    def forward(self, x):
        x = self.fc(x)
        return x

class GAP(nn.Module):
    def __init__(self, kernel):
        super(GAP, self).__init__()
        self.k = kernel
        # self.fc = nn.Linear(512, 1024)
    def forward(self, x):
        x = F.adaptive_avg_pool2d(x, self.k)

        return x.squeeze(-1).squeeze(-1)

class AdaptiveSpatialFeatureFusion(nn.Module):
    def __init__(self, args, in_channels=[512, 1024, 1024],
                 out_channels=[256, 512, 1024]):

        super(AdaptiveSpatialFeatureFusion, self).__init__()
        self.weight = nn.LayerNorm(out_channels[2])
        self.proj = linear_layer(args, in_channels[0], out_channels[2])

    def forward(self, feature_map1, feature_map2):
        # feature_map1 : b, 1024, 1, 1
        # feature_map2 : b, 512, 1, 1
        feature_map2 = self.proj(feature_map2.squeeze(-1).squeeze(-1))
        feature_map1 = feature_map1.squeeze(-1).squeeze(-1)
        weights1 = torch.norm(feature_map1, dim=1).unsqueeze(-1)
        weights2 = torch.norm(feature_map2, dim=1).unsqueeze(-1)
        weights1 = weights1 / (weights1 + weights2)
        weights2 = 1 - weights1

        fused_feature_map = weights1 * feature_map1 + weights2 * feature_map2
        # b, 1024
        return fused_feature_map

class ModifiedAttentionPool2d(nn.Module):
    def __init__(self,
                 spacial_dim: int,
                 embed_dim: int,
                 num_heads: int,
                 output_dim: int = None):
        super().__init__()
        self.spacial_dim = spacial_dim
        self.positional_embedding = nn.Parameter(
            torch.randn(spacial_dim**2 + 1, embed_dim) / embed_dim**0.5)
        self.k_proj = nn.Linear(embed_dim, embed_dim)
        self.q_proj = nn.Linear(embed_dim, embed_dim)
        self.v_proj = nn.Linear(embed_dim, embed_dim)
        self.c_proj = nn.Linear(embed_dim, output_dim or embed_dim)
        self.num_heads = num_heads
        # residual
        self.connect = nn.Sequential(
            nn.Conv2d(embed_dim, output_dim, 1, stride=1, bias=False),
            nn.BatchNorm2d(output_dim))

    def resize_pos_embed(self, pos_embed, input_shpae):
        """Resize pos_embed weights.
        Resize pos_embed using bicubic interpolate method.
        Args:
            pos_embed (torch.Tensor): Position embedding weights.
            input_shpae (tuple): Tuple for (downsampled input image height,
                downsampled input image width).
            pos_shape (tuple): The resolution of downsampled origin training
                image.
            mode (str): Algorithm used for upsampling:
                ``'nearest'`` | ``'linear'`` | ``'bilinear'`` | ``'bicubic'`` |
                ``'trilinear'``. Default: ``'nearest'``
        Return:
            torch.Tensor: The resized pos_embed of shape [B, C, L_new]
        """
        assert pos_embed.ndim == 3, 'shape of pos_embed must be [B, L, C]'
        pos_h = pos_w = self.spacial_dim
        cls_token_weight = pos_embed[:, 0]
        pos_embed_weight = pos_embed[:, (-1 * pos_h * pos_w):]
        pos_embed_weight = pos_embed_weight.reshape(
            1, pos_h, pos_w, pos_embed.shape[2]).permute(0, 3, 1, 2)
        pos_embed_weight = F.interpolate(pos_embed_weight,
                                         size=input_shpae,
                                         align_corners=False,
                                         mode='bicubic')
        cls_token_weight = cls_token_weight.unsqueeze(1)
        pos_embed_weight = torch.flatten(pos_embed_weight, 2).transpose(1, 2)
        # pos_embed = torch.cat((cls_token_weight, pos_embed_weight), dim=1)
        return pos_embed_weight.transpose(-2, -1)

    def forward(self, x):
        B, C, H, W = x.size()
        res = self.connect(x)
        x = x.reshape(B, C, -1)  # NC(HW)
        # x = torch.cat([x.mean(dim=-1, keepdim=True), x], dim=-1)  # NC(1+HW)
        pos_embed = self.positional_embedding.unsqueeze(0)
        pos_embed = self.resize_pos_embed(pos_embed, (H, W))  # NC(HW)
        x = x + pos_embed.to(x.dtype)  # NC(HW)
        x = x.permute(2, 0, 1)  # (HW)NC
        x, _ = F.multi_head_attention_forward(
            query=x,
            key=x,
            value=x,
            embed_dim_to_check=x.shape[-1],
            num_heads=self.num_heads,
            q_proj_weight=self.q_proj.weight,
            k_proj_weight=self.k_proj.weight,
            v_proj_weight=self.v_proj.weight,
            in_proj_weight=None,
            in_proj_bias=torch.cat(
                [self.q_proj.bias, self.k_proj.bias, self.v_proj.bias]),
            bias_k=None,
            bias_v=None,
            add_zero_attn=False,
            dropout_p=0,
            out_proj_weight=self.c_proj.weight,
            out_proj_bias=self.c_proj.bias,
            use_separate_proj_weight=True,
            training=self.training,
            need_weights=False)
        xt = x[0]
        x = x.permute(1, 2, 0).reshape(B, -1, H, W)
        x = x + res
        x = F.relu(x, True)

        return x, xt

# modified
class FPN(nn.Module):
    def __init__(self, args,
                 in_channels=[512, 1024, 1024],
                 out_channels=[256, 512, 1024, 1024]):
        super(FPN, self).__init__()
        input_resolution = args.input_size
        heads = args.heads
        output_dim = args.output_dim
        embed_dim = args.emb_dim
        # image projection
        self.attn = ModifiedAttentionPool2d(input_resolution // 32, embed_dim, heads, output_dim)
        # text projection
        self.txt_proj = linear_layer(args, in_channels[2], out_channels[2])
        # fusion 1: v5 & seq -> f_5: b, 1024, 13, 13
        self.f1_v_proj = conv_layer(in_channels[2], out_channels[2], 1, 0)

        self.norm_layer = nn.Sequential(nn.BatchNorm2d(out_channels[2]),
                                        nn.ReLU(True))

        # fusion 2: v4 & fm -> f_4: b, 512, 26, 26
        self.f2_v_proj = conv_layer(in_channels[1], out_channels[1], 3, 1)
        self.f2_cat = conv_layer(out_channels[2] + out_channels[1],
                                 out_channels[1], 1, 0)
        # fusion 3: v3 & fm_mid -> f_3: b, 512, 52, 52
        self.f3_v_proj = conv_layer(in_channels[0], out_channels[0], 3, 1)
        self.f3_cat = conv_layer(out_channels[0] + out_channels[1],
                                 out_channels[1], 1, 0)
        # fusion 4: f_3 & f_4 & f_5 -> fq: b, 256, 26, 26
        self.f4_proj5 = conv_layer(out_channels[2], out_channels[1], 3, 1)
        self.f4_proj4 = conv_layer(out_channels[1], out_channels[1], 3, 1)
        self.f4_proj3 = conv_layer(out_channels[1], out_channels[1], 3, 1)
        # aggregation
        self.aggr = conv_layer(3 * out_channels[1], out_channels[1], 1, 0)
        self.coordconv = nn.Sequential(
            CoordConv(out_channels[1], out_channels[1], 3, 1),
            conv_layer(out_channels[1], out_channels[3], 3, 1))

    def forward(self, imgs, text):
        # v3, v4, v5: 256, 52, 52 / 512, 26, 26 / 1024, 13, 13
        v3, v4, v5 = imgs

        # fusion 1: b, 1024, 13, 13
        # text projection: b, 1024 -> b, 1024
        v5, _ = self.attn(v5)
        text_ = self.txt_proj(text)
        state = text_.unsqueeze(-1).unsqueeze(
            -1)# b, 1024, 1, 1

        f5 = self.f1_v_proj(v5) # b, 1024, 7, 7

        f5 = self.norm_layer(f5 * state)
        # fusion 2: b, 512, 26, 26
        f4 = self.f2_v_proj(v4)
        # f4 = f4.repeat(w2,1,1,1)

        f5_ = F.interpolate(f5, scale_factor=2, mode='bilinear')
        f4 = self.f2_cat(torch.cat([f4, f5_], dim=1))
        # fusion 3: b, 256, 26, 26
        f3 = self.f3_v_proj(v3)
        f3 = F.avg_pool2d(f3, 2, 2)
        # f3 = f3.repeat(w2, 1, 1, 1)

        f3 = self.f3_cat(torch.cat([f3, f4], dim=1))
        # fusion 4: b, 512, 13, 13 / b, 512, 26, 26 / b, 512, 26, 26
        fq5 = self.f4_proj5(f5)
        fq4 = self.f4_proj4(f4)
        fq3 = self.f4_proj3(f3)
        # query
        fq5 = F.interpolate(fq5, scale_factor=2, mode='bilinear')
        fq = torch.cat([fq3, fq4, fq5], dim=1)
        fq = self.aggr(fq)
        fq = self.coordconv(fq)
            # fqq = fq.reshape(w1, w2, fq.shape[1], fq.shape[2], fq.shape[3])
            # b, 512, 26, 26

        # elif text.shape[0] != v3.shape[0]:
        #
        #     text = self.txt_proj(text)
        #     state = text.unsqueeze(-1).unsqueeze(
        #         -1)  # b, 1024, 1, 1
        #     state = state.view(v5.shape[0], int(text.shape[0] / v5.shape[0]), state.shape[1], state.shape[2], state.shape[3])
        #
        #     f5 = self.f1_v_proj(v5)  # b, 1024, 7, 7
        #     f5 = f5.unsqueeze(1)
        #     f5_ = f5 * state
        #     f5_ = f5_.view(-1, f5.shape[2], f5.shape[3], f5.shape[4])
        #     f5 = self.norm_layer(f5_)
        #     # fusion 2: b, 512, 26, 26
        #     f4 = self.f2_v_proj(v4)
        #     # f4 = f4.repeat(w2,1,1,1)
        #
        #     f5_ = F.interpolate(f5, scale_factor=2, mode='bilinear')
        #     f4 = f4.repeat(int(f5_.shape[0] / f4.shape[0]), 1, 1, 1)
        #     f4 = self.f2_cat(torch.cat([f4, f5_], dim=1))
        #
        #     # fusion 3: b, 256, 26, 26
        #     f3 = self.f3_v_proj(v3)
        #     f3 = F.avg_pool2d(f3, 2, 2)
        #     # f3 = f3.repeat(w2, 1, 1, 1)
        #     f3 = f3.repeat(int(f5_.shape[0] / f3.shape[0]), 1, 1, 1)
        #     f3 = self.f3_cat(torch.cat([f3, f4], dim=1))
        #     # fusion 4: b, 512, 13, 13 / b, 512, 26, 26 / b, 512, 26, 26
        #     fq5 = self.f4_proj5(f5)
        #     fq4 = self.f4_proj4(f4)
        #     fq3 = self.f4_proj3(f3)
        #     # query
        #     fq5 = F.interpolate(fq5, scale_factor=2, mode='bilinear')
        #     fq = torch.cat([fq3, fq4, fq5], dim=1)
        #     fq = self.aggr(fq)
        #     fq = self.coordconv(fq)
        return fq

class ViTFPN(nn.Module):
    def __init__(self, image_resolution,
                 in_channels=[512, 768, 768],
                 out_channels=[768, 768, 768, 512]):
        super(ViTFPN, self).__init__()
        # text projection
        self.txt_proj = linear_layer(in_channels[0], out_channels[1])
        # fusion 1: v5 & seq -> f_5: b, 1024, 13, 13
        self.f1_v_proj = conv_layer(in_channels[1], out_channels[1], 1, 0)
        self.norm_layer = nn.Sequential(nn.BatchNorm2d(out_channels[1]),
                                        nn.ReLU(True))
        # fusion 2: v4 & fm -> f_4: b, 512, 26, 26
        self.f2_v_proj = conv_layer(in_channels[1], out_channels[1], 3, 1)
        self.f2_cat = conv_layer(out_channels[0] + out_channels[0],
                                 out_channels[0], 1, 0)
        # fusion 3: v3 & fm_mid -> f_3: b, 512, 52, 52
        self.f3_v_proj = conv_layer(in_channels[1], out_channels[1], 3, 1)
        self.f3_cat = conv_layer(out_channels[0] + out_channels[1],
                                 out_channels[1], 1, 0)
        # fusion 4: f_3 & f_4 & f_5 -> fq: b, 256, 26, 26
        self.f4_proj5 = conv_layer(out_channels[1], out_channels[0], 3, 1)
        self.f4_proj4 = conv_layer(out_channels[0], out_channels[0], 3, 1)
        self.f4_proj3 = conv_layer(out_channels[1], out_channels[1], 3, 1)
        # aggregation
        self.aggr = conv_layer(3 * out_channels[0], out_channels[0], 1, 0)
        self.coordconv = nn.Sequential(
            CoordConv(out_channels[0], out_channels[0], 3, 1),
            conv_layer(out_channels[0], out_channels[-1], 3, 1))

        self.attnpool = AttentionPool2d(image_resolution // 32, out_channels[-1],
                                    8, out_channels[-1])
    def forward(self, imgs, state, vis):
        # v1 / v2 / b, 49, 1024/ b, 196, 512
        v3, v4, v5 = imgs
        # fusion 1: b, 1024, 13, 13
        # text projection: b, 1024 -> b, 1024
        state = self.txt_proj(state)
        state = state.unsqueeze(-1).unsqueeze(
            -1)# b, 1024, 1, 1
        f5 = self.f1_v_proj(v5)
        f5 = self.norm_layer(f5 * state)
        # fusion 2: b, 512, 26, 26
        f4 = self.f2_v_proj(v4)
        b, c, h, w = f4.size()
        f5_ = F.interpolate(f5, (h, w), mode='bilinear')
        f4 = self.f2_cat(torch.cat([f4, f5_], dim=1))

        # fusion 3: b, 256, 26, 26
        f3 = self.f3_v_proj(v3)
        f3 = F.avg_pool2d(f3, 2, 2)
        # f3 = f3.repeat(w2, 1, 1, 1)

        f3 = self.f3_cat(torch.cat([f3, f4], dim=1))
        # fusion 4: b, 512, 13, 13 / b, 512, 26, 26 / b, 512, 26, 26
        fq5 = self.f4_proj5(f5)
        fq4 = self.f4_proj4(f4)
        fq3 = self.f4_proj3(f3)
        # query
        fq5 = F.interpolate(fq5, (h, w), mode='bilinear')
        fq = torch.cat([fq3, fq4, fq5], dim=1)
        fq = self.aggr(fq)
        if not vis:
            fq = self.coordconv(fq)
            fq = self.attnpool(fq)
        # b, 512, 26, 26
        return fq