Add app.py, requirements.txt, examples and model files 2

lib/pvt.py

@@ -0,0 +1,223 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from lib.pvtv2 import pvt_v2_b2
+import os
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class BasicConv2d(nn.Module):
+    def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1):
+        super(BasicConv2d, self).__init__()
+
+        self.conv = nn.Conv2d(in_planes, out_planes,
+                              kernel_size=kernel_size, stride=stride,
+                              padding=padding, dilation=dilation, bias=False)
+        self.bn = nn.BatchNorm2d(out_planes)
+        self.relu = nn.ReLU(inplace=True)
+
+    def forward(self, x):
+        x = self.conv(x)
+        x = self.bn(x)
+        return x
+
+
+class CFM(nn.Module):
+    def __init__(self, channel):
+        super(CFM, self).__init__()
+        self.relu = nn.ReLU(True)
+
+        self.upsample = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)
+        self.conv_upsample1 = BasicConv2d(channel, channel, 3, padding=1)
+        self.conv_upsample2 = BasicConv2d(channel, channel, 3, padding=1)
+        self.conv_upsample3 = BasicConv2d(channel, channel, 3, padding=1)
+        self.conv_upsample4 = BasicConv2d(channel, channel, 3, padding=1)
+        self.conv_upsample5 = BasicConv2d(2 * channel, 2 * channel, 3, padding=1)
+
+        self.conv_concat2 = BasicConv2d(2 * channel, 2 * channel, 3, padding=1)
+        self.conv_concat3 = BasicConv2d(3 * channel, 3 * channel, 3, padding=1)
+        self.conv4 = BasicConv2d(3 * channel, channel, 3, padding=1)
+
+    def forward(self, x1, x2, x3):
+        x1_1 = x1
+        x2_1 = self.conv_upsample1(self.upsample(x1)) * x2
+        x3_1 = self.conv_upsample2(self.upsample(self.upsample(x1))) \
+               * self.conv_upsample3(self.upsample(x2)) * x3
+
+        x2_2 = torch.cat((x2_1, self.conv_upsample4(self.upsample(x1_1))), 1)
+        x2_2 = self.conv_concat2(x2_2)
+
+        x3_2 = torch.cat((x3_1, self.conv_upsample5(self.upsample(x2_2))), 1)
+        x3_2 = self.conv_concat3(x3_2)
+
+        x1 = self.conv4(x3_2)
+
+        return x1
+
+
+
+
+class GCN(nn.Module):
+    def __init__(self, num_state, num_node, bias=False):
+        super(GCN, self).__init__()
+        self.conv1 = nn.Conv1d(num_node, num_node, kernel_size=1)
+        self.relu = nn.ReLU(inplace=True)
+        self.conv2 = nn.Conv1d(num_state, num_state, kernel_size=1, bias=bias)
+
+    def forward(self, x):
+        h = self.conv1(x.permute(0, 2, 1)).permute(0, 2, 1)
+        h = h - x
+        h = self.relu(self.conv2(h))
+        return h
+
+
+class SAM(nn.Module):
+    def __init__(self, num_in=32, plane_mid=16, mids=4, normalize=False):
+        super(SAM, self).__init__()
+
+        self.normalize = normalize
+        self.num_s = int(plane_mid)
+        self.num_n = (mids) * (mids)
+        self.priors = nn.AdaptiveAvgPool2d(output_size=(mids + 2, mids + 2))
+
+        self.conv_state = nn.Conv2d(num_in, self.num_s, kernel_size=1)
+        self.conv_proj = nn.Conv2d(num_in, self.num_s, kernel_size=1)
+        self.gcn = GCN(num_state=self.num_s, num_node=self.num_n)
+        self.conv_extend = nn.Conv2d(self.num_s, num_in, kernel_size=1, bias=False)
+
+    def forward(self, x, edge):
+        edge = F.upsample(edge, (x.size()[-2], x.size()[-1]))
+
+        n, c, h, w = x.size()
+        edge = torch.nn.functional.softmax(edge, dim=1)[:, 1, :, :].unsqueeze(1)
+
+        x_state_reshaped = self.conv_state(x).view(n, self.num_s, -1)
+        x_proj = self.conv_proj(x)
+        x_mask = x_proj * edge
+
+        x_anchor1 = self.priors(x_mask)
+        x_anchor2 = self.priors(x_mask)[:, :, 1:-1, 1:-1].reshape(n, self.num_s, -1)
+        x_anchor = self.priors(x_mask)[:, :, 1:-1, 1:-1].reshape(n, self.num_s, -1)
+
+        x_proj_reshaped = torch.matmul(x_anchor.permute(0, 2, 1), x_proj.reshape(n, self.num_s, -1))
+        x_proj_reshaped = torch.nn.functional.softmax(x_proj_reshaped, dim=1)
+
+        x_rproj_reshaped = x_proj_reshaped
+
+        x_n_state = torch.matmul(x_state_reshaped, x_proj_reshaped.permute(0, 2, 1))
+        if self.normalize:
+            x_n_state = x_n_state * (1. / x_state_reshaped.size(2))
+        x_n_rel = self.gcn(x_n_state)
+
+        x_state_reshaped = torch.matmul(x_n_rel, x_rproj_reshaped)
+        x_state = x_state_reshaped.view(n, self.num_s, *x.size()[2:])
+        out = x + (self.conv_extend(x_state))
+
+        return out
+
+
+class ChannelAttention(nn.Module):
+    def __init__(self, in_planes, ratio=16):
+        super(ChannelAttention, self).__init__()
+        self.avg_pool = nn.AdaptiveAvgPool2d(1)
+        self.max_pool = nn.AdaptiveMaxPool2d(1)
+
+        self.fc1   = nn.Conv2d(in_planes, in_planes // 16, 1, bias=False)
+        self.relu1 = nn.ReLU()
+        self.fc2   = nn.Conv2d(in_planes // 16, in_planes, 1, bias=False)
+
+        self.sigmoid = nn.Sigmoid()
+
+    def forward(self, x):
+        avg_out = self.fc2(self.relu1(self.fc1(self.avg_pool(x))))
+        max_out = self.fc2(self.relu1(self.fc1(self.max_pool(x))))
+        out = avg_out + max_out
+        return self.sigmoid(out)
+
+
+class SpatialAttention(nn.Module):
+    def __init__(self, kernel_size=7):
+        super(SpatialAttention, self).__init__()
+
+        assert kernel_size in (3, 7), 'kernel size must be 3 or 7'
+        padding = 3 if kernel_size == 7 else 1
+
+        self.conv1 = nn.Conv2d(2, 1, kernel_size, padding=padding, bias=False)
+        self.sigmoid = nn.Sigmoid()
+
+    def forward(self, x):
+        avg_out = torch.mean(x, dim=1, keepdim=True)
+        max_out, _ = torch.max(x, dim=1, keepdim=True)
+        x = torch.cat([avg_out, max_out], dim=1)
+        x = self.conv1(x)
+        return self.sigmoid(x)
+
+
+class PolypPVT(nn.Module):
+    def __init__(self, channel=32):
+        super(PolypPVT, self).__init__()
+
+        self.backbone = pvt_v2_b2()  # [64, 128, 320, 512]
+        path = './pretrained_pth/pvt_v2_b2.pth'
+        save_model = torch.load(path)
+        model_dict = self.backbone.state_dict()
+        state_dict = {k: v for k, v in save_model.items() if k in model_dict.keys()}
+        model_dict.update(state_dict)
+        self.backbone.load_state_dict(model_dict)
+
+        self.Translayer2_0 = BasicConv2d(64, channel, 1)
+        self.Translayer2_1 = BasicConv2d(128, channel, 1)
+        self.Translayer3_1 = BasicConv2d(320, channel, 1)
+        self.Translayer4_1 = BasicConv2d(512, channel, 1)
+
+        self.CFM = CFM(channel)
+        self.ca = ChannelAttention(64)
+        self.sa = SpatialAttention()
+        self.SAM = SAM()
+        
+        self.down05 = nn.Upsample(scale_factor=0.5, mode='bilinear', align_corners=True)
+        self.out_SAM = nn.Conv2d(channel, 1, 1)
+        self.out_CFM = nn.Conv2d(channel, 1, 1)
+
+
+    def forward(self, x):
+
+        # backbone
+        pvt = self.backbone(x)
+        x1 = pvt[0]
+        x2 = pvt[1]
+        x3 = pvt[2]
+        x4 = pvt[3]
+        
+        # CIM
+        x1 = self.ca(x1) * x1 # channel attention
+        cim_feature = self.sa(x1) * x1 # spatial attention
+
+
+        # CFM
+        x2_t = self.Translayer2_1(x2)  
+        x3_t = self.Translayer3_1(x3)  
+        x4_t = self.Translayer4_1(x4)  
+        cfm_feature = self.CFM(x4_t, x3_t, x2_t)
+
+        # SAM
+        T2 = self.Translayer2_0(cim_feature)
+        T2 = self.down05(T2)
+        sam_feature = self.SAM(cfm_feature, T2)
+
+        prediction1 = self.out_CFM(cfm_feature)
+        prediction2 = self.out_SAM(sam_feature)
+
+        prediction1_8 = F.interpolate(prediction1, scale_factor=8, mode='bilinear') 
+        prediction2_8 = F.interpolate(prediction2, scale_factor=8, mode='bilinear')  
+        return prediction1_8, prediction2_8
+
+
+if __name__ == '__main__':
+    model = PolypPVT().cuda()
+    input_tensor = torch.randn(1, 3, 352, 352).cuda()
+
+    prediction1, prediction2 = model(input_tensor)
+    print(prediction1.size(), prediction2.size())

lib/pvtv2.py

@@ -0,0 +1,436 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from functools import partial
+
+from timm.models.layers import DropPath, to_2tuple, trunc_normal_
+from timm.models.registry import register_model
+from timm.models.vision_transformer import _cfg
+from timm.models.registry import register_model
+
+import math
+
+
+class Mlp(nn.Module):
+    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.fc1 = nn.Linear(in_features, hidden_features)
+        self.dwconv = DWConv(hidden_features)
+        self.act = act_layer()
+        self.fc2 = nn.Linear(hidden_features, out_features)
+        self.drop = nn.Dropout(drop)
+
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+        elif isinstance(m, nn.Conv2d):
+            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+            fan_out //= m.groups
+            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
+            if m.bias is not None:
+                m.bias.data.zero_()
+
+    def forward(self, x, H, W):
+        x = self.fc1(x)
+        x = self.dwconv(x, H, W)
+        x = self.act(x)
+        x = self.drop(x)
+        x = self.fc2(x)
+        x = self.drop(x)
+        return x
+
+
+class Attention(nn.Module):
+    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0., sr_ratio=1):
+        super().__init__()
+        assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}."
+
+        self.dim = dim
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        self.scale = qk_scale or head_dim ** -0.5
+
+        self.q = nn.Linear(dim, dim, bias=qkv_bias)
+        self.kv = nn.Linear(dim, dim * 2, bias=qkv_bias)
+        self.attn_drop = nn.Dropout(attn_drop)
+        self.proj = nn.Linear(dim, dim)
+        self.proj_drop = nn.Dropout(proj_drop)
+
+        self.sr_ratio = sr_ratio
+        if sr_ratio > 1:
+            self.sr = nn.Conv2d(dim, dim, kernel_size=sr_ratio, stride=sr_ratio)
+            self.norm = nn.LayerNorm(dim)
+
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+        elif isinstance(m, nn.Conv2d):
+            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+            fan_out //= m.groups
+            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
+            if m.bias is not None:
+                m.bias.data.zero_()
+
+    def forward(self, x, H, W):
+        B, N, C = x.shape
+        q = self.q(x).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)
+
+        if self.sr_ratio > 1:
+            x_ = x.permute(0, 2, 1).reshape(B, C, H, W)
+            x_ = self.sr(x_).reshape(B, C, -1).permute(0, 2, 1)
+            x_ = self.norm(x_)
+            kv = self.kv(x_).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
+        else:
+            kv = self.kv(x).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
+        k, v = kv[0], kv[1]
+
+        attn = (q @ k.transpose(-2, -1)) * self.scale
+        attn = attn.softmax(dim=-1)
+        attn = self.attn_drop(attn)
+
+        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
+        x = self.proj(x)
+        x = self.proj_drop(x)
+
+        return x
+
+
+class Block(nn.Module):
+
+    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
+                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm, sr_ratio=1):
+        super().__init__()
+        self.norm1 = norm_layer(dim)
+        self.attn = Attention(
+            dim,
+            num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,
+            attn_drop=attn_drop, proj_drop=drop, sr_ratio=sr_ratio)
+        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
+        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
+        self.norm2 = norm_layer(dim)
+        mlp_hidden_dim = int(dim * mlp_ratio)
+        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)
+
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+        elif isinstance(m, nn.Conv2d):
+            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+            fan_out //= m.groups
+            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
+            if m.bias is not None:
+                m.bias.data.zero_()
+
+    def forward(self, x, H, W):
+        x = x + self.drop_path(self.attn(self.norm1(x), H, W))
+        x = x + self.drop_path(self.mlp(self.norm2(x), H, W))
+
+        return x
+
+
+class OverlapPatchEmbed(nn.Module):
+    """ Image to Patch Embedding
+    """
+
+    def __init__(self, img_size=224, patch_size=7, stride=4, in_chans=3, embed_dim=768):
+        super().__init__()
+        img_size = to_2tuple(img_size)
+        patch_size = to_2tuple(patch_size)
+
+        self.img_size = img_size
+        self.patch_size = patch_size
+        self.H, self.W = img_size[0] // patch_size[0], img_size[1] // patch_size[1]
+        self.num_patches = self.H * self.W
+        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=stride,
+                              padding=(patch_size[0] // 2, patch_size[1] // 2))
+        self.norm = nn.LayerNorm(embed_dim)
+
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+        elif isinstance(m, nn.Conv2d):
+            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+            fan_out //= m.groups
+            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
+            if m.bias is not None:
+                m.bias.data.zero_()
+
+    def forward(self, x):
+        x = self.proj(x)
+        _, _, H, W = x.shape
+        x = x.flatten(2).transpose(1, 2)
+        x = self.norm(x)
+
+        return x, H, W
+
+
+class PyramidVisionTransformerImpr(nn.Module):
+    def __init__(self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, embed_dims=[64, 128, 256, 512],
+                 num_heads=[1, 2, 4, 8], mlp_ratios=[4, 4, 4, 4], qkv_bias=False, qk_scale=None, drop_rate=0.,
+                 attn_drop_rate=0., drop_path_rate=0., norm_layer=nn.LayerNorm,
+                 depths=[3, 4, 6, 3], sr_ratios=[8, 4, 2, 1]):
+        super().__init__()
+        self.num_classes = num_classes
+        self.depths = depths
+
+        # patch_embed
+        self.patch_embed1 = OverlapPatchEmbed(img_size=img_size, patch_size=7, stride=4, in_chans=in_chans,
+                                              embed_dim=embed_dims[0])
+        self.patch_embed2 = OverlapPatchEmbed(img_size=img_size // 4, patch_size=3, stride=2, in_chans=embed_dims[0],
+                                              embed_dim=embed_dims[1])
+        self.patch_embed3 = OverlapPatchEmbed(img_size=img_size // 8, patch_size=3, stride=2, in_chans=embed_dims[1],
+                                              embed_dim=embed_dims[2])
+        self.patch_embed4 = OverlapPatchEmbed(img_size=img_size // 16, patch_size=3, stride=2, in_chans=embed_dims[2],
+                                              embed_dim=embed_dims[3])
+
+        # transformer encoder
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule
+        cur = 0
+        self.block1 = nn.ModuleList([Block(
+            dim=embed_dims[0], num_heads=num_heads[0], mlp_ratio=mlp_ratios[0], qkv_bias=qkv_bias, qk_scale=qk_scale,
+            drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[cur + i], norm_layer=norm_layer,
+            sr_ratio=sr_ratios[0])
+            for i in range(depths[0])])
+        self.norm1 = norm_layer(embed_dims[0])
+
+        cur += depths[0]
+        self.block2 = nn.ModuleList([Block(
+            dim=embed_dims[1], num_heads=num_heads[1], mlp_ratio=mlp_ratios[1], qkv_bias=qkv_bias, qk_scale=qk_scale,
+            drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[cur + i], norm_layer=norm_layer,
+            sr_ratio=sr_ratios[1])
+            for i in range(depths[1])])
+        self.norm2 = norm_layer(embed_dims[1])
+
+        cur += depths[1]
+        self.block3 = nn.ModuleList([Block(
+            dim=embed_dims[2], num_heads=num_heads[2], mlp_ratio=mlp_ratios[2], qkv_bias=qkv_bias, qk_scale=qk_scale,
+            drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[cur + i], norm_layer=norm_layer,
+            sr_ratio=sr_ratios[2])
+            for i in range(depths[2])])
+        self.norm3 = norm_layer(embed_dims[2])
+
+        cur += depths[2]
+        self.block4 = nn.ModuleList([Block(
+            dim=embed_dims[3], num_heads=num_heads[3], mlp_ratio=mlp_ratios[3], qkv_bias=qkv_bias, qk_scale=qk_scale,
+            drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[cur + i], norm_layer=norm_layer,
+            sr_ratio=sr_ratios[3])
+            for i in range(depths[3])])
+        self.norm4 = norm_layer(embed_dims[3])
+
+        # classification head
+        # self.head = nn.Linear(embed_dims[3], num_classes) if num_classes > 0 else nn.Identity()
+
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+        elif isinstance(m, nn.Conv2d):
+            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
+            fan_out //= m.groups
+            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
+            if m.bias is not None:
+                m.bias.data.zero_()
+
+    def init_weights(self, pretrained=None):
+        if isinstance(pretrained, str):
+            logger = 1
+            #load_checkpoint(self, pretrained, map_location='cpu', strict=False, logger=logger)
+
+    def reset_drop_path(self, drop_path_rate):
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(self.depths))]
+        cur = 0
+        for i in range(self.depths[0]):
+            self.block1[i].drop_path.drop_prob = dpr[cur + i]
+
+        cur += self.depths[0]
+        for i in range(self.depths[1]):
+            self.block2[i].drop_path.drop_prob = dpr[cur + i]
+
+        cur += self.depths[1]
+        for i in range(self.depths[2]):
+            self.block3[i].drop_path.drop_prob = dpr[cur + i]
+
+        cur += self.depths[2]
+        for i in range(self.depths[3]):
+            self.block4[i].drop_path.drop_prob = dpr[cur + i]
+
+    def freeze_patch_emb(self):
+        self.patch_embed1.requires_grad = False
+
+    @torch.jit.ignore
+    def no_weight_decay(self):
+        return {'pos_embed1', 'pos_embed2', 'pos_embed3', 'pos_embed4', 'cls_token'}  # has pos_embed may be better
+
+    def get_classifier(self):
+        return self.head
+
+    def reset_classifier(self, num_classes, global_pool=''):
+        self.num_classes = num_classes
+        self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()
+
+    # def _get_pos_embed(self, pos_embed, patch_embed, H, W):
+    #     if H * W == self.patch_embed1.num_patches:
+    #         return pos_embed
+    #     else:
+    #         return F.interpolate(
+    #             pos_embed.reshape(1, patch_embed.H, patch_embed.W, -1).permute(0, 3, 1, 2),
+    #             size=(H, W), mode="bilinear").reshape(1, -1, H * W).permute(0, 2, 1)
+
+    def forward_features(self, x):
+        B = x.shape[0]
+        outs = []
+
+        # stage 1
+        x, H, W = self.patch_embed1(x)
+        for i, blk in enumerate(self.block1):
+            x = blk(x, H, W)
+        x = self.norm1(x)
+        x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
+        outs.append(x)
+
+        # stage 2
+        x, H, W = self.patch_embed2(x)
+        for i, blk in enumerate(self.block2):
+            x = blk(x, H, W)
+        x = self.norm2(x)
+        x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
+        outs.append(x)
+
+        # stage 3
+        x, H, W = self.patch_embed3(x)
+        for i, blk in enumerate(self.block3):
+            x = blk(x, H, W)
+        x = self.norm3(x)
+        x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
+        outs.append(x)
+
+        # stage 4
+        x, H, W = self.patch_embed4(x)
+        for i, blk in enumerate(self.block4):
+            x = blk(x, H, W)
+        x = self.norm4(x)
+        x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
+        outs.append(x)
+
+        return outs
+
+        # return x.mean(dim=1)
+
+    def forward(self, x):
+        x = self.forward_features(x)
+        # x = self.head(x)
+
+        return x
+
+
+class DWConv(nn.Module):
+    def __init__(self, dim=768):
+        super(DWConv, self).__init__()
+        self.dwconv = nn.Conv2d(dim, dim, 3, 1, 1, bias=True, groups=dim)
+
+    def forward(self, x, H, W):
+        B, N, C = x.shape
+        x = x.transpose(1, 2).view(B, C, H, W)
+        x = self.dwconv(x)
+        x = x.flatten(2).transpose(1, 2)
+
+        return x
+
+
+def _conv_filter(state_dict, patch_size=16):
+    """ convert patch embedding weight from manual patchify + linear proj to conv"""
+    out_dict = {}
+    for k, v in state_dict.items():
+        if 'patch_embed.proj.weight' in k:
+            v = v.reshape((v.shape[0], 3, patch_size, patch_size))
+        out_dict[k] = v
+
+    return out_dict
+
+
+@register_model
+class pvt_v2_b0(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b0, self).__init__(
+            patch_size=4, embed_dims=[32, 64, 160, 256], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[2, 2, 2, 2], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)
+
+
+
+@register_model
+class pvt_v2_b1(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b1, self).__init__(
+            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[2, 2, 2, 2], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)
+
+@register_model
+class pvt_v2_b2(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b2, self).__init__(
+            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 4, 6, 3], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)
+
+@register_model
+class pvt_v2_b3(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b3, self).__init__(
+            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 4, 18, 3], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)
+
+@register_model
+class pvt_v2_b4(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b4, self).__init__(
+            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[8, 8, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 8, 27, 3], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)
+
+
+@register_model
+class pvt_v2_b5(PyramidVisionTransformerImpr):
+    def __init__(self, **kwargs):
+        super(pvt_v2_b5, self).__init__(
+            patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[4, 4, 4, 4],
+            qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6), depths=[3, 6, 40, 3], sr_ratios=[8, 4, 2, 1],
+            drop_rate=0.0, drop_path_rate=0.1)