UNetFormer_MMSAM_heatmap.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from einops import rearrange, repeat
import numpy as np
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
import timm
import cv2
import torch.autograd as autograd
from MedSAM.models.sam import sam_model_registry
import MedSAM.cfg as cfg
import matplotlib.pyplot as plt

class Norm2d(nn.Module):
    def __init__(self, embed_dim):
        super().__init__()
        self.ln = nn.LayerNorm(embed_dim, eps=1e-6)
    def forward(self, x):
        x = x.permute(0, 2, 3, 1)
        x = self.ln(x)
        x = x.permute(0, 3, 1, 2).contiguous()
        return x
    
class ConvBNReLU(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, dilation=1, stride=1, norm_layer=nn.BatchNorm2d, bias=False):
        super(ConvBNReLU, self).__init__(
            nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size, bias=bias,
                      dilation=dilation, stride=stride, padding=((stride - 1) + dilation * (kernel_size - 1)) // 2),
            norm_layer(out_channels),
            nn.ReLU6()
        )


class ConvBN(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, dilation=1, stride=1, norm_layer=nn.BatchNorm2d, bias=False):
        super(ConvBN, self).__init__(
            nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size, bias=bias,
                      dilation=dilation, stride=stride, padding=((stride - 1) + dilation * (kernel_size - 1)) // 2),
            norm_layer(out_channels)
        )


class Conv(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, dilation=1, stride=1, bias=False):
        super(Conv, self).__init__(
            nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size, bias=bias,
                      dilation=dilation, stride=stride, padding=((stride - 1) + dilation * (kernel_size - 1)) // 2)
        )


class SeparableConvBNReLU(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, dilation=1,
                 norm_layer=nn.BatchNorm2d):
        super(SeparableConvBNReLU, self).__init__(
            nn.Conv2d(in_channels, in_channels, kernel_size, stride=stride, dilation=dilation,
                      padding=((stride - 1) + dilation * (kernel_size - 1)) // 2,
                      groups=in_channels, bias=False),
            norm_layer(out_channels),
            nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False),
            nn.ReLU6()
        )


class SeparableConvBN(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, dilation=1,
                 norm_layer=nn.BatchNorm2d):
        super(SeparableConvBN, self).__init__(
            nn.Conv2d(in_channels, in_channels, kernel_size, stride=stride, dilation=dilation,
                      padding=((stride - 1) + dilation * (kernel_size - 1)) // 2,
                      groups=in_channels, bias=False),
            norm_layer(out_channels),
            nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
        )


class SeparableConv(nn.Sequential):
    def __init__(self, in_channels, out_channels, kernel_size=3, stride=1, dilation=1):
        super(SeparableConv, self).__init__(
            nn.Conv2d(in_channels, in_channels, kernel_size, stride=stride, dilation=dilation,
                      padding=((stride - 1) + dilation * (kernel_size - 1)) // 2,
                      groups=in_channels, bias=False),
            nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
        )


class Mlp(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.ReLU6, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Conv2d(in_features, hidden_features, 1, 1, 0, bias=True)
        self.act = act_layer()
        self.fc2 = nn.Conv2d(hidden_features, out_features, 1, 1, 0, bias=True)
        self.drop = nn.Dropout(drop, inplace=True)

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop(x)
        x = self.fc2(x)
        x = self.drop(x)
        return x


class GlobalLocalAttention(nn.Module):
    def __init__(self,
                 dim=256,
                 num_heads=16,
                 qkv_bias=False,
                 window_size=8,
                 relative_pos_embedding=True
                 ):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // self.num_heads
        self.scale = head_dim ** -0.5
        self.ws = window_size

        self.qkv = Conv(dim, 3*dim, kernel_size=1, bias=qkv_bias)
        self.local1 = ConvBN(dim, dim, kernel_size=3)
        self.local2 = ConvBN(dim, dim, kernel_size=1)
        self.proj = SeparableConvBN(dim, dim, kernel_size=window_size)

        self.attn_x = nn.AvgPool2d(kernel_size=(window_size, 1), stride=1,  padding=(window_size//2 - 1, 0))
        self.attn_y = nn.AvgPool2d(kernel_size=(1, window_size), stride=1, padding=(0, window_size//2 - 1))

        self.relative_pos_embedding = relative_pos_embedding

        if self.relative_pos_embedding:
            # define a parameter table of relative position bias
            self.relative_position_bias_table = nn.Parameter(
                torch.zeros((2 * window_size - 1) * (2 * window_size - 1), num_heads))  # 2*Wh-1 * 2*Ww-1, nH

            # get pair-wise relative position index for each token inside the window
            coords_h = torch.arange(self.ws)
            coords_w = torch.arange(self.ws)
            coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
            coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
            relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
            relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
            relative_coords[:, :, 0] += self.ws - 1  # shift to start from 0
            relative_coords[:, :, 1] += self.ws - 1
            relative_coords[:, :, 0] *= 2 * self.ws - 1
            relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
            self.register_buffer("relative_position_index", relative_position_index)

            trunc_normal_(self.relative_position_bias_table, std=.02)

    def pad(self, x, ps):
        _, _, H, W = x.size()
        if W % ps != 0:
            x = F.pad(x, (0, ps - W % ps), mode='reflect')
        if H % ps != 0:
            x = F.pad(x, (0, 0, 0, ps - H % ps), mode='reflect')
        return x

    def pad_out(self, x):
        x = F.pad(x, pad=(0, 1, 0, 1), mode='reflect')
        return x

    def forward(self, x):
        B, C, H, W = x.shape

        local = self.local2(x) + self.local1(x)

        x = self.pad(x, self.ws)
        B, C, Hp, Wp = x.shape
        qkv = self.qkv(x)

        q, k, v = rearrange(qkv, 'b (qkv h d) (hh ws1) (ww ws2) -> qkv (b hh ww) h (ws1 ws2) d', h=self.num_heads,
                            d=C//self.num_heads, hh=Hp//self.ws, ww=Wp//self.ws, qkv=3, ws1=self.ws, ws2=self.ws)

        dots = (q @ k.transpose(-2, -1)) * self.scale

        if self.relative_pos_embedding:
            relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
                self.ws * self.ws, self.ws * self.ws, -1)  # Wh*Ww,Wh*Ww,nH
            relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
            dots += relative_position_bias.unsqueeze(0)

        attn = dots.softmax(dim=-1)
        attn = attn @ v

        attn = rearrange(attn, '(b hh ww) h (ws1 ws2) d -> b (h d) (hh ws1) (ww ws2)', h=self.num_heads,
                         d=C//self.num_heads, hh=Hp//self.ws, ww=Wp//self.ws, ws1=self.ws, ws2=self.ws)

        attn = attn[:, :, :H, :W]

        out = self.attn_x(F.pad(attn, pad=(0, 0, 0, 1), mode='reflect')) + \
              self.attn_y(F.pad(attn, pad=(0, 1, 0, 0), mode='reflect'))

        out = out + local
        out = self.pad_out(out)
        out = self.proj(out)
        # print(out.size())
        out = out[:, :, :H, :W]

        return out

class Block(nn.Module):
    def __init__(self, dim=256, num_heads=16,  mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0.,
                 drop_path=0., act_layer=nn.ReLU6, norm_layer=nn.BatchNorm2d, window_size=8):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = GlobalLocalAttention(dim, num_heads=num_heads, qkv_bias=qkv_bias, window_size=window_size)

        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, out_features=dim, act_layer=act_layer, drop=drop)
        self.norm2 = norm_layer(dim)

    def forward(self, x):

        x = x + self.drop_path(self.attn(self.norm1(x)))
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x

class WF_single(nn.Module):
    def __init__(self, in_channels=128, decode_channels=128, eps=1e-8):
        super(WF_single, self).__init__()
        self.pre_conv = Conv(in_channels, decode_channels, kernel_size=1)

        self.weights = nn.Parameter(torch.ones(2, dtype=torch.float32), requires_grad=True)
        self.eps = eps
        self.post_conv = ConvBNReLU(decode_channels, decode_channels, kernel_size=3)

    def forward(self, x):
        x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=False)
        x = self.post_conv(x)
        return x
    
class WF(nn.Module):
    def __init__(self, in_channels=128, decode_channels=128, eps=1e-8):
        super(WF, self).__init__()
        self.pre_conv = Conv(in_channels, decode_channels, kernel_size=1)

        self.weights = nn.Parameter(torch.ones(2, dtype=torch.float32), requires_grad=True)
        self.eps = eps
        self.post_conv = ConvBNReLU(decode_channels, decode_channels, kernel_size=3)

    def forward(self, x, res):
        x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=False)
        weights = nn.ReLU()(self.weights)
        fuse_weights = weights / (torch.sum(weights, dim=0) + self.eps)
        x = fuse_weights[0] * self.pre_conv(res) + fuse_weights[1] * x
        x = self.post_conv(x)
        return x


class SqueezeAndExcitation(nn.Module):
    def __init__(self, channel,
                 reduction=16, activation=nn.ReLU(inplace=True)):
        super(SqueezeAndExcitation, self).__init__()
        self.fc = nn.Sequential(
            nn.Conv2d(channel, channel // reduction, kernel_size=1),
            activation,
            nn.Conv2d(channel // reduction, channel, kernel_size=1),
            nn.Sigmoid()
        )

    def forward(self, x):
        weighting = F.adaptive_avg_pool2d(x, 1)
        weighting = self.fc(weighting)
        y = x * weighting
        return y
    
class SEFusion(nn.Module):
    def __init__(self, channels_in, activation=nn.ReLU(inplace=True)):
        super(SEFusion, self).__init__()

        self.se_rgb = SqueezeAndExcitation(channels_in,
                                           activation=activation)
        self.se_depth = SqueezeAndExcitation(channels_in,
                                             activation=activation)

    def forward(self, rgb, depth):
        rgb = self.se_rgb(rgb)
        depth = self.se_depth(depth)
        out = rgb + depth
        return out

  
class FeatureRefinementHead_single(nn.Module):
    def __init__(self, in_channels=64, decode_channels=64):
        super().__init__()
        self.pre_conv = Conv(in_channels, decode_channels, kernel_size=1)

        self.weights = nn.Parameter(torch.ones(2, dtype=torch.float32), requires_grad=True)
        self.eps = 1e-8
        self.post_conv = ConvBNReLU(decode_channels, decode_channels, kernel_size=3)

        self.pa = nn.Sequential(nn.Conv2d(decode_channels, decode_channels, kernel_size=3, padding=1, groups=decode_channels),
                                nn.Sigmoid())
        self.ca = nn.Sequential(nn.AdaptiveAvgPool2d(1),
                                Conv(decode_channels, decode_channels//16, kernel_size=1),
                                nn.ReLU6(),
                                Conv(decode_channels//16, decode_channels, kernel_size=1),
                                nn.Sigmoid())

        self.shortcut = ConvBN(decode_channels, decode_channels, kernel_size=1)
        self.proj = SeparableConvBN(decode_channels, decode_channels, kernel_size=3)
        self.act = nn.ReLU6()

    def forward(self, x):
        x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=False)
        # weights = nn.ReLU()(self.weights)
        # fuse_weights = weights / (torch.sum(weights, dim=0) + self.eps)
        # x = fuse_weights[0] * self.pre_conv(res) + fuse_weights[1] * x
        x = self.post_conv(x)
        shortcut = self.shortcut(x)
        pa = self.pa(x) * x
        ca = self.ca(x) * x
        x = pa + ca
        x = self.proj(x) + shortcut
        x = self.act(x)

        return x
      
class FeatureRefinementHead(nn.Module):
    def __init__(self, in_channels=64, decode_channels=64):
        super().__init__()
        self.pre_conv = Conv(in_channels, decode_channels, kernel_size=1)

        self.weights = nn.Parameter(torch.ones(2, dtype=torch.float32), requires_grad=True)
        self.eps = 1e-8
        self.post_conv = ConvBNReLU(decode_channels, decode_channels, kernel_size=3)

        self.pa = nn.Sequential(nn.Conv2d(decode_channels, decode_channels, kernel_size=3, padding=1, groups=decode_channels),
                                nn.Sigmoid())
        self.ca = nn.Sequential(nn.AdaptiveAvgPool2d(1),
                                Conv(decode_channels, decode_channels//16, kernel_size=1),
                                nn.ReLU6(),
                                Conv(decode_channels//16, decode_channels, kernel_size=1),
                                nn.Sigmoid())

        self.shortcut = ConvBN(decode_channels, decode_channels, kernel_size=1)
        self.proj = SeparableConvBN(decode_channels, decode_channels, kernel_size=3)
        self.act = nn.ReLU6()

    def forward(self, x, res):
        x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=False)
        weights = nn.ReLU()(self.weights)
        fuse_weights = weights / (torch.sum(weights, dim=0) + self.eps)
        x = fuse_weights[0] * self.pre_conv(res) + fuse_weights[1] * x
        x = self.post_conv(x)
        shortcut = self.shortcut(x)
        pa = self.pa(x) * x
        ca = self.ca(x) * x
        x = pa + ca
        x = self.proj(x) + shortcut
        x = self.act(x)

        return x

class AuxHead(nn.Module):

    def __init__(self, in_channels=64, num_classes=8):
        super().__init__()
        self.conv = ConvBNReLU(in_channels, in_channels)
        self.drop = nn.Dropout(0.1)
        self.conv_out = Conv(in_channels, num_classes, kernel_size=1)

    def forward(self, x, h, w):
        feat = self.conv(x)
        feat = self.drop(feat)
        feat = self.conv_out(feat)
        feat = F.interpolate(feat, size=(h, w), mode='bilinear', align_corners=False)
        return feat


class Decoder_single(nn.Module):
    def __init__(self,
                 encoder_channels=(64, 128, 256, 512),
                 decode_channels=64,
                 dropout=0.1,
                 window_size=8,
                 num_classes=6):
        super(Decoder_single, self).__init__()

        self.pre_conv = ConvBN(encoder_channels[-1], decode_channels, kernel_size=1)
        self.b4 = Block(dim=decode_channels, num_heads=8, window_size=window_size)

        self.b3 = Block(dim=decode_channels, num_heads=8, window_size=window_size)
        self.p3 = WF_single(encoder_channels[-2], decode_channels)

        self.b2 = Block(dim=decode_channels, num_heads=8, window_size=window_size)
        self.p2 = WF_single(encoder_channels[-3], decode_channels)

        self.p1 = FeatureRefinementHead_single(encoder_channels[-4], decode_channels)

        self.segmentation_head = nn.Sequential(ConvBNReLU(decode_channels, decode_channels),
                                               nn.Dropout2d(p=dropout, inplace=True),
                                               Conv(decode_channels, num_classes, kernel_size=1))
        self.init_weight()

    def forward(self, res4, h, w):
        x = self.b4(self.pre_conv(res4))
        x = self.p3(x)
        x = self.b3(x)

        x = self.p2(x)
        x = self.b2(x)

        x = self.p1(x)
        
        x = self.segmentation_head(x)
        x = F.interpolate(x, size=(h, w), mode='bilinear', align_corners=False)

        return x

    def init_weight(self):
        for m in self.children():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, a=1)
                if m.bias is not None:
                    nn.init.constant_(m.bias, 0)
                    
class Decoder(nn.Module):
    def __init__(self,
                 encoder_channels=(64, 128, 256, 512),
                 decode_channels=64,
                 dropout=0.1,
                 window_size=8,
                 num_classes=6):
        super(Decoder, self).__init__()

        self.pre_conv = ConvBN(encoder_channels[-1], decode_channels, kernel_size=1)
        self.b4 = Block(dim=decode_channels, num_heads=8, window_size=window_size)

        self.b3 = Block(dim=decode_channels, num_heads=8, window_size=window_size)
        self.p3 = WF(encoder_channels[-2], decode_channels)

        self.b2 = Block(dim=decode_channels, num_heads=8, window_size=window_size)
        self.p2 = WF(encoder_channels[-3], decode_channels)

        self.p1 = FeatureRefinementHead(encoder_channels[-4], decode_channels)

        self.segmentation_head = nn.Sequential(ConvBNReLU(decode_channels, decode_channels),
                                               nn.Dropout2d(p=dropout, inplace=True),
                                               Conv(decode_channels, num_classes, kernel_size=1))
        self.init_weight()

    def forward(self, res1, res2, res3, res4, h, w):
        x = self.b4(self.pre_conv(res4))
        x = self.p3(x, res3)
        x = self.b3(x)

        x = self.p2(x, res2)
        x = self.b2(x)

        x = self.p1(x, res1)
        
        x = self.segmentation_head(x)
        x = F.interpolate(x, size=(h, w), mode='bilinear', align_corners=False)

        return x

    def init_weight(self):
        for m in self.children():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, a=1)
                if m.bias is not None:
                    nn.init.constant_(m.bias, 0)

def draw_features(feature, savename=''):
    H = W = 256
    visualize = F.interpolate(feature, size=(H, W), mode='bilinear', align_corners=False)
    visualize = visualize.detach().cpu().numpy()
    visualize = np.mean(visualize, axis=1).reshape(H, W)
    visualize = (((visualize - np.min(visualize)) / (np.max(visualize) - np.min(visualize))) * 255).astype(np.uint8)
    # fvis = np.fft.fft2(visualize)
    # fshift = np.fft.fftshift(fvis)
    # fshift = 20*np.log(np.abs(fshift))
    savedir = savename
    visualize = cv2.applyColorMap(visualize, cv2.COLORMAP_JET)
    cv2.imwrite(savedir, visualize)
    
class UNetFormer(nn.Module):
    def __init__(self,
                 decode_channels=64,
                 dropout=0.1,
                 window_size=8,
                 num_classes=6
                 ):
        super().__init__()
        args = cfg.parse_args()
        # self.sam = sam_model_registry["vit_b"](args,checkpoint='weights/sam_vit_b_01ec64.pth')
        self.sam = sam_model_registry["vit_l"](args,checkpoint='weights/sam_vit_l_0b3195.pth')
        # self.sam = sam_model_registry["vit_h"](args,checkpoint='weights/sam_vit_h_4b8939.pth')
        self.image_encoder = self.sam.image_encoder
        encoder_channels = (256, 256, 256, 256)
        
        self.fpn1x = nn.Sequential(
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
            Norm2d(256),
            nn.GELU(),
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
        )
        self.fpn2x = nn.Sequential(
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
        )
        self.fpn3x = nn.Identity()
        self.fpn4x = nn.MaxPool2d(kernel_size=2, stride=2)
        
        self.fpn1y = nn.Sequential(
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
            nn.BatchNorm2d(256),
            nn.GELU(),
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
        )
        self.fpn2y = nn.Sequential(
            nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2),
        )
        self.fpn3y = nn.Identity()
        self.fpn4y = nn.MaxPool2d(kernel_size=2, stride=2)
        
        self.fusion1 = SEFusion(encoder_channels[0])
        self.fusion2 = SEFusion(encoder_channels[1])
        self.fusion3 = SEFusion(encoder_channels[2])
        self.fusion4 = SEFusion(encoder_channels[3])
        for n, value in self.image_encoder.named_parameters(): 
            # if "Adapter" not in n:
            if 'lora_' not in n:
            # if 'lora_' not in n and "Adapter" not in n:
                value.requires_grad = False
            else:
                value.requires_grad = True

        self.decoder = Decoder(encoder_channels, decode_channels, dropout, window_size, num_classes)
        # self.decoder = Decoder_single(encoder_channels, decode_channels, dropout, window_size, num_classes)

    def forward(self, x, y, mode='Train'):
        h, w = x.size()[-2:]
        y = torch.unsqueeze(y, dim=1).repeat(1,3,1,1)
        deepx, deepy = self.image_encoder(x, y) # 256*16*16
        heatmaps = []
        heatmaps.append(deepx)
        heatmaps.append(deepy)
        # #PFF:
        res1x = self.fpn1x(deepx)
        res2x = self.fpn2x(deepx)
        res3x = self.fpn3x(deepx)
        res4x = self.fpn4x(deepx)
        res1y = self.fpn1y(deepy)
        res2y = self.fpn2y(deepy)
        res3y = self.fpn3y(deepy)
        res4y = self.fpn4y(deepy)
        res1 = self.fusion1(res1x, res1y)
        res2 = self.fusion2(res2x, res2y)
        res3 = self.fusion3(res3x, res3y)
        res4 = self.fusion4(res4x, res4y)

        x = self.decoder(res1, res2, res3, res4, h, w)
        
        # ## without PFF:
        # res4 = deepx + deepy
        # x = self.decoder(res4, h, w)
        pred = x[:, 1, 100, 65]
        ## heatmap
        feature = heatmaps[0]
        feature_grad = autograd.grad(pred, feature, allow_unused=True, retain_graph=True)[0]
        grads = feature_grad  # 获取梯度
        pooled_grads = torch.nn.functional.adaptive_avg_pool2d(grads, (1, 1))
        # 此处batch size默认为1，所以去掉了第0维（batch size维）
        pooled_grads = pooled_grads[0]
        feature = feature[0]
        # print("pooled_grads:", pooled_grads.shape)
        # print("feature:", feature.shape)
        # feature.shape[0]是指定层feature的通道数
        for i in range(feature.shape[0]):
            feature[i, ...] *= pooled_grads[i, ...]
        heatmap = feature.detach().cpu().numpy()
        heatmap = np.mean(heatmap, axis=0)
        heatmap1 = np.maximum(heatmap, 0)
        heatmap1 /= np.max(heatmap1)
        
        feature = heatmaps[1]
        feature_grad = autograd.grad(pred, feature, allow_unused=True, retain_graph=True)[0]
        grads = feature_grad  # 获取梯度
        pooled_grads = torch.nn.functional.adaptive_avg_pool2d(grads, (1, 1))
        # 此处batch size默认为1，所以去掉了第0维（batch size维）
        pooled_grads = pooled_grads[0]
        feature = feature[0]
        # print("pooled_grads:", pooled_grads.shape)
        # print("feature:", feature.shape)
        # feature.shape[0]是指定层feature的通道数
        for i in range(feature.shape[0]):
            feature[i, ...] *= pooled_grads[i, ...]
        heatmap = feature.detach().cpu().numpy()
        heatmap = np.mean(heatmap, axis=0)
        heatmap2 = np.maximum(heatmap, 0)
        heatmap2 /= np.max(heatmap2)
        return x, heatmap1, heatmap2