基于Attention U-Net的宠物图像分割

本文基于《Attention U-Net: Learning Where to Look for the Pancreas》，实现了用于宠物图像分割的Attention U-Net模型。通过划分数据集，构建含注意力门的网络结构，用RMSProp优化器和交叉熵损失训练，经15轮后在测试集上预测，结果展示了模型对宠物图像的分割效果，验证了其有效性。

基于Attention U-Net的宠物图像分割

论文：Attention U-Net: Learning Where to Look for the Pancreas

简介

效果

模型结构

Attention Gate模块

Attention的意思是，把注意力放到目标区域上，简单来说就是让目标区域的值变大。Attention模块用在了skip connection上，原始U-Net只是单纯的把同层的下采样层的特征直接concate到上采样层中，改进后的使用attention模块对下采样层同层和上采样层上一层的特征图进行处理后再和上采样后的特征图进行concate

环境设置

import osimport ioimport numpy as npimport matplotlib.pyplot as pltfrom PIL import Image as PilImageimport paddleimport paddle.nn as nnimport paddle.nn.functional as Fpaddle.set_device('gpu')paddle.__version__

'2.1.0'

数据处理

此处数据处理部分借鉴了『跟着雨哥学AI』系列06：趣味案例——基于U-Net的宠物图像分割

# 解压缩!tar -xf data/data50154/images.tar.gz!tar -xf data/data50154/annotations.tar.gz

IMAGE_SIZE = (160, 160)train_images_path = "images/"label_images_path = "annotations/trimaps/"image_count = len([os.path.join(train_images_path, image_name)           for image_name in os.listdir(train_images_path)           if image_name.endswith('.webp')])print("用于训练的图片样本数量:", image_count)# 对数据集进行处理，划分训练集、测试集def _sort_images(image_dir, image_type):    """    对文件夹内的图像进行按照文件名排序    """    files = []    for image_name in os.listdir(image_dir):        if image_name.endswith('.{}'.format(image_type)) \                and not image_name.startswith('.'):            files.append(os.path.join(image_dir, image_name))    return sorted(files)def write_file(mode, images, labels):    with open('./{}.txt'.format(mode), 'w') as f:        for i in range(len(images)):            f.write('{}\t{}\n'.format(images[i], labels[i]))    images = _sort_images(train_images_path, 'jpg')labels = _sort_images(label_images_path, 'png')eval_num = int(image_count * 0.15)write_file('train', images[:-eval_num], labels[:-eval_num])write_file('test', images[-eval_num:], labels[-eval_num:])write_file('predict', images[-eval_num:], labels[-eval_num:])

用于训练的图片样本数量: 7390

with open('./train.txt', 'r') as f:    i = 0    for line in f.readlines():        image_path, label_path = line.strip().split('\t')        image = np.array(PilImage.open(image_path))        label = np.array(PilImage.open(label_path))            if i > 2:            break        # 进行图片的展示        plt.figure()        plt.subplot(1,2,1),         plt.title('Train Image')        plt.imshow(image.astype('uint8'))        plt.axis('off')        plt.subplot(1,2,2),         plt.title('Label')        plt.imshow(label.astype('uint8'), cmap='gray')        plt.axis('off')        plt.show()        i = i + 1

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2349: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working  if isinstance(obj, collections.Iterator):/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2366: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working  return list(data) if isinstance(data, collections.MappingView) else data/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/image.py:425: DeprecationWarning: np.asscalar(a) is deprecated since NumPy v1.16, use a.item() instead  a_min = np.asscalar(a_min.astype(scaled_dtype))/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/image.py:426: DeprecationWarning: np.asscalar(a) is deprecated since NumPy v1.16, use a.item() instead  a_max = np.asscalar(a_max.astype(scaled_dtype))

数据集类定义

import randomfrom paddle.io import Datasetfrom paddle.vision.transforms import transforms as Tclass PetDataset(Dataset):    """    数据集定义    """    def __init__(self, mode='train'):        """        构造函数        """        self.image_size = IMAGE_SIZE        self.mode = mode.lower()                assert self.mode in ['train', 'test', 'predict'], \            "mode should be 'train' or 'test' or 'predict', but got {}".format(self.mode)                self.train_images = []        self.label_images = []        with open('./{}.txt'.format(self.mode), 'r') as f:            for line in f.readlines():                image, label = line.strip().split('\t')                self.train_images.append(image)                self.label_images.append(label)            def _load_img(self, path, color_mode='rgb', transforms=[]):        """        统一的图像处理接口封装，用于规整图像大小和通道        """        with open(path, 'rb') as f:            img = PilImage.open(io.BytesIO(f.read()))            if color_mode == 'grayscale':                # if image is not already an 8-bit, 16-bit or 32-bit grayscale image                # convert it to an 8-bit grayscale image.                if img.mode not in ('L', 'I;16', 'I'):                    img = img.convert('L')            elif color_mode == 'rgba':                if img.mode != 'RGBA':                    img = img.convert('RGBA')            elif color_mode == 'rgb':                if img.mode != 'RGB':                    img = img.convert('RGB')            else:                raise ValueError('color_mode must be "grayscale", "rgb", or "rgba"')                        return T.Compose([                T.Resize(self.image_size)            ] + transforms)(img)    def __getitem__(self, idx):        """        返回 image, label        """        train_image = self._load_img(self.train_images[idx],                                      transforms=[                                         T.Transpose(),                                          T.Normalize(mean=127.5, std=127.5)                                     ]) # 加载原始图像        label_image = self._load_img(self.label_images[idx],                                      color_mode='grayscale',                                     transforms=[T.Grayscale()]) # 加载Label图像            # 返回image, label        train_image = np.array(train_image, dtype='float32')        label_image = np.array(label_image, dtype='int64')        return train_image, label_image            def __len__(self):        """        返回数据集总数        """        return len(self.train_images)

模型组网

基础模块

class conv_block(nn.Layer):    def __init__(self, ch_in, ch_out):        super(conv_block, self).__init__()        self.conv = nn.Sequential(            nn.Conv2D(ch_in, ch_out, kernel_size=3, stride=1, padding=1),            nn.BatchNorm(ch_out),            nn.ReLU(),            nn.Conv2D(ch_out, ch_out, kernel_size=3, stride=1, padding=1),            nn.BatchNorm(ch_out),            nn.ReLU()        )    def forward(self, x):        x = self.conv(x)        return xclass up_conv(nn.Layer):    def __init__(self, ch_in, ch_out):        super(up_conv, self).__init__()        self.up = nn.Sequential(            nn.Upsample(scale_factor=2),            nn.Conv2D(ch_in, ch_out, kernel_size=3, stride=1, padding=1),            nn.BatchNorm(ch_out),            nn.ReLU()        )    def forward(self, x):        x = self.up(x)        return xclass single_conv(nn.Layer):    def __init__(self, ch_in, ch_out):        super(single_conv, self).__init__()        self.conv = nn.Sequential(            nn.Conv2D(ch_in, ch_out, kernel_size=3, stride=1, padding=1),            nn.BatchNorm(ch_out),            nn.ReLU()        )    def forward(self, x):        x = self.conv(x)        return x

Attention块

class Attention_block(nn.Layer):    def __init__(self, F_g, F_l, F_int):        super(Attention_block, self).__init__()        self.W_g = nn.Sequential(            nn.Conv2D(F_g, F_int, kernel_size=1, stride=1, padding=0),            nn.BatchNorm(F_int)        )        self.W_x = nn.Sequential(            nn.Conv2D(F_l, F_int, kernel_size=1, stride=1, padding=0),            nn.BatchNorm(F_int)        )        self.psi = nn.Sequential(            nn.Conv2D(F_int, 1, kernel_size=1, stride=1, padding=0),            nn.BatchNorm(1),            nn.Sigmoid()        )        self.relu = nn.ReLU()    def forward(self, g, x):        g1 = self.W_g(g)        x1 = self.W_x(x)        psi = self.relu(g1 + x1)        psi = self.psi(psi)        return x * psi

Attention U-Net

class AttU_Net(nn.Layer):    def __init__(self, img_ch=3, output_ch=1):        super(AttU_Net, self).__init__()        self.Maxpool = nn.MaxPool2D(kernel_size=2, stride=2)        self.Maxpool1 = nn.MaxPool2D(kernel_size=2, stride=2)        self.Maxpool2 = nn.MaxPool2D(kernel_size=2, stride=2)        self.Maxpool3 = nn.MaxPool2D(kernel_size=2, stride=2)        self.Conv1 = conv_block(ch_in=img_ch, ch_out=64)        self.Conv2 = conv_block(ch_in=64, ch_out=128)        self.Conv3 = conv_block(ch_in=128, ch_out=256)        self.Conv4 = conv_block(ch_in=256, ch_out=512)        self.Conv5 = conv_block(ch_in=512, ch_out=1024)        self.Up5 = up_conv(ch_in=1024, ch_out=512)        self.Att5 = Attention_block(F_g=512, F_l=512, F_int=256)        self.Up_conv5 = conv_block(ch_in=1024, ch_out=512)        self.Up4 = up_conv(ch_in=512, ch_out=256)        self.Att4 = Attention_block(F_g=256, F_l=256, F_int=128)        self.Up_conv4 = conv_block(ch_in=512, ch_out=256)        self.Up3 = up_conv(ch_in=256, ch_out=128)        self.Att3 = Attention_block(F_g=128, F_l=128, F_int=64)        self.Up_conv3 = conv_block(ch_in=256, ch_out=128)        self.Up2 = up_conv(ch_in=128, ch_out=64)        self.Att2 = Attention_block(F_g=64, F_l=64, F_int=32)        self.Up_conv2 = conv_block(ch_in=128, ch_out=64)        self.Conv_1x1 = nn.Conv2D(64, output_ch, kernel_size=1, stride=1, padding=0)    def forward(self, x):        # encoding path        x1 = self.Conv1(x)        x2 = self.Maxpool(x1)        x2 = self.Conv2(x2)        x3 = self.Maxpool1(x2)        x3 = self.Conv3(x3)        x4 = self.Maxpool2(x3)        x4 = self.Conv4(x4)        x5 = self.Maxpool3(x4)        x5 = self.Conv5(x5)        # decoding + concat path        d5 = self.Up5(x5)        x4 = self.Att5(g=d5, x=x4)        d5 = paddle.concat(x=[x4, d5], axis=1)        d5 = self.Up_conv5(d5)        d4 = self.Up4(d5)        x3 = self.Att4(g=d4, x=x3)        d4 = paddle.concat(x=[x3, d4], axis=1)        d4 = self.Up_conv4(d4)        d3 = self.Up3(d4)        x2 = self.Att3(g=d3, x=x2)        d3 = paddle.concat(x=[x2, d3], axis=1)        d3 = self.Up_conv3(d3)        d2 = self.Up2(d3)        x1 = self.Att2(g=d2, x=x1)        d2 = paddle.concat(x=[x1, d2], axis=1)        d2 = self.Up_conv2(d2)        d1 = self.Conv_1x1(d2)        return d1

模型可视化

num_classes = 4network = AttU_Net(img_ch=3, output_ch=num_classes)model = paddle.Model(network)model.summary((-1, 3,) + IMAGE_SIZE)

-----------------------------------------------------------------------------  Layer (type)        Input Shape          Output Shape         Param #    =============================================================================    Conv2D-1       [[1, 3, 160, 160]]   [1, 64, 160, 160]        1,792        BatchNorm-1    [[1, 64, 160, 160]]   [1, 64, 160, 160]         256           ReLU-1       [[1, 64, 160, 160]]   [1, 64, 160, 160]          0           Conv2D-2      [[1, 64, 160, 160]]   [1, 64, 160, 160]       36,928        BatchNorm-2    [[1, 64, 160, 160]]   [1, 64, 160, 160]         256           ReLU-2       [[1, 64, 160, 160]]   [1, 64, 160, 160]          0         conv_block-1     [[1, 3, 160, 160]]   [1, 64, 160, 160]          0          MaxPool2D-1    [[1, 64, 160, 160]]    [1, 64, 80, 80]           0           Conv2D-3       [[1, 64, 80, 80]]     [1, 128, 80, 80]       73,856        BatchNorm-3     [[1, 128, 80, 80]]    [1, 128, 80, 80]         512           ReLU-3        [[1, 128, 80, 80]]    [1, 128, 80, 80]          0           Conv2D-4       [[1, 128, 80, 80]]    [1, 128, 80, 80]       147,584       BatchNorm-4     [[1, 128, 80, 80]]    [1, 128, 80, 80]         512           ReLU-4        [[1, 128, 80, 80]]    [1, 128, 80, 80]          0         conv_block-2     [[1, 64, 80, 80]]     [1, 128, 80, 80]          0          MaxPool2D-2     [[1, 128, 80, 80]]    [1, 128, 40, 40]          0           Conv2D-5       [[1, 128, 40, 40]]    [1, 256, 40, 40]       295,168       BatchNorm-5     [[1, 256, 40, 40]]    [1, 256, 40, 40]        1,024          ReLU-5        [[1, 256, 40, 40]]    [1, 256, 40, 40]          0           Conv2D-6       [[1, 256, 40, 40]]    [1, 256, 40, 40]       590,080       BatchNorm-6     [[1, 256, 40, 40]]    [1, 256, 40, 40]        1,024          ReLU-6        [[1, 256, 40, 40]]    [1, 256, 40, 40]          0         conv_block-3     [[1, 128, 40, 40]]    [1, 256, 40, 40]          0          MaxPool2D-3     [[1, 256, 40, 40]]    [1, 256, 20, 20]          0           Conv2D-7       [[1, 256, 20, 20]]    [1, 512, 20, 20]      1,180,160      BatchNorm-7     [[1, 512, 20, 20]]    [1, 512, 20, 20]        2,048          ReLU-7        [[1, 512, 20, 20]]    [1, 512, 20, 20]          0           Conv2D-8       [[1, 512, 20, 20]]    [1, 512, 20, 20]      2,359,808      BatchNorm-8     [[1, 512, 20, 20]]    [1, 512, 20, 20]        2,048          ReLU-8        [[1, 512, 20, 20]]    [1, 512, 20, 20]          0         conv_block-4     [[1, 256, 20, 20]]    [1, 512, 20, 20]          0          MaxPool2D-4     [[1, 512, 20, 20]]    [1, 512, 10, 10]          0           Conv2D-9       [[1, 512, 10, 10]]   [1, 1024, 10, 10]      4,719,616      BatchNorm-9    [[1, 1024, 10, 10]]   [1, 1024, 10, 10]        4,096          ReLU-9       [[1, 1024, 10, 10]]   [1, 1024, 10, 10]          0           Conv2D-10     [[1, 1024, 10, 10]]   [1, 1024, 10, 10]      9,438,208     BatchNorm-10    [[1, 1024, 10, 10]]   [1, 1024, 10, 10]        4,096          ReLU-10      [[1, 1024, 10, 10]]   [1, 1024, 10, 10]          0         conv_block-5     [[1, 512, 10, 10]]   [1, 1024, 10, 10]          0          Upsample-1     [[1, 1024, 10, 10]]   [1, 1024, 20, 20]          0           Conv2D-11     [[1, 1024, 20, 20]]    [1, 512, 20, 20]      4,719,104     BatchNorm-11     [[1, 512, 20, 20]]    [1, 512, 20, 20]        2,048          ReLU-11       [[1, 512, 20, 20]]    [1, 512, 20, 20]          0           up_conv-1     [[1, 1024, 10, 10]]    [1, 512, 20, 20]          0           Conv2D-12      [[1, 512, 20, 20]]    [1, 256, 20, 20]       131,328      BatchNorm-12     [[1, 256, 20, 20]]    [1, 256, 20, 20]        1,024         Conv2D-13      [[1, 512, 20, 20]]    [1, 256, 20, 20]       131,328      BatchNorm-13     [[1, 256, 20, 20]]    [1, 256, 20, 20]        1,024          ReLU-12       [[1, 256, 20, 20]]    [1, 256, 20, 20]          0           Conv2D-14      [[1, 256, 20, 20]]     [1, 1, 20, 20]          257        BatchNorm-14      [[1, 1, 20, 20]]      [1, 1, 20, 20]           4           Sigmoid-1       [[1, 1, 20, 20]]      [1, 1, 20, 20]           0       Attention_block-1          []            [1, 512, 20, 20]          0           Conv2D-15     [[1, 1024, 20, 20]]    [1, 512, 20, 20]      4,719,104     BatchNorm-15     [[1, 512, 20, 20]]    [1, 512, 20, 20]        2,048          ReLU-13       [[1, 512, 20, 20]]    [1, 512, 20, 20]          0           Conv2D-16      [[1, 512, 20, 20]]    [1, 512, 20, 20]      2,359,808     BatchNorm-16     [[1, 512, 20, 20]]    [1, 512, 20, 20]        2,048          ReLU-14       [[1, 512, 20, 20]]    [1, 512, 20, 20]          0         conv_block-6    [[1, 1024, 20, 20]]    [1, 512, 20, 20]          0          Upsample-2      [[1, 512, 20, 20]]    [1, 512, 40, 40]          0           Conv2D-17      [[1, 512, 40, 40]]    [1, 256, 40, 40]      1,179,904     BatchNorm-17     [[1, 256, 40, 40]]    [1, 256, 40, 40]        1,024          ReLU-15       [[1, 256, 40, 40]]    [1, 256, 40, 40]          0           up_conv-2      [[1, 512, 20, 20]]    [1, 256, 40, 40]          0           Conv2D-18      [[1, 256, 40, 40]]    [1, 128, 40, 40]       32,896       BatchNorm-18     [[1, 128, 40, 40]]    [1, 128, 40, 40]         512          Conv2D-19      [[1, 256, 40, 40]]    [1, 128, 40, 40]       32,896       BatchNorm-19     [[1, 128, 40, 40]]    [1, 128, 40, 40]         512           ReLU-16       [[1, 128, 40, 40]]    [1, 128, 40, 40]          0           Conv2D-20      [[1, 128, 40, 40]]     [1, 1, 40, 40]          129        BatchNorm-20      [[1, 1, 40, 40]]      [1, 1, 40, 40]           4           Sigmoid-2       [[1, 1, 40, 40]]      [1, 1, 40, 40]           0       Attention_block-2          []            [1, 256, 40, 40]          0           Conv2D-21      [[1, 512, 40, 40]]    [1, 256, 40, 40]      1,179,904     BatchNorm-21     [[1, 256, 40, 40]]    [1, 256, 40, 40]        1,024          ReLU-17       [[1, 256, 40, 40]]    [1, 256, 40, 40]          0           Conv2D-22      [[1, 256, 40, 40]]    [1, 256, 40, 40]       590,080      BatchNorm-22     [[1, 256, 40, 40]]    [1, 256, 40, 40]        1,024          ReLU-18       [[1, 256, 40, 40]]    [1, 256, 40, 40]          0         conv_block-7     [[1, 512, 40, 40]]    [1, 256, 40, 40]          0          Upsample-3      [[1, 256, 40, 40]]    [1, 256, 80, 80]          0           Conv2D-23      [[1, 256, 80, 80]]    [1, 128, 80, 80]       295,040      BatchNorm-23     [[1, 128, 80, 80]]    [1, 128, 80, 80]         512           ReLU-19       [[1, 128, 80, 80]]    [1, 128, 80, 80]          0           up_conv-3      [[1, 256, 40, 40]]    [1, 128, 80, 80]          0           Conv2D-24      [[1, 128, 80, 80]]    [1, 64, 80, 80]         8,256       BatchNorm-24     [[1, 64, 80, 80]]     [1, 64, 80, 80]          256          Conv2D-25      [[1, 128, 80, 80]]    [1, 64, 80, 80]         8,256       BatchNorm-25     [[1, 64, 80, 80]]     [1, 64, 80, 80]          256           ReLU-20       [[1, 64, 80, 80]]     [1, 64, 80, 80]           0           Conv2D-26      [[1, 64, 80, 80]]      [1, 1, 80, 80]          65         BatchNorm-26      [[1, 1, 80, 80]]      [1, 1, 80, 80]           4           Sigmoid-3       [[1, 1, 80, 80]]      [1, 1, 80, 80]           0       Attention_block-3          []            [1, 128, 80, 80]          0           Conv2D-27      [[1, 256, 80, 80]]    [1, 128, 80, 80]       295,040      BatchNorm-27     [[1, 128, 80, 80]]    [1, 128, 80, 80]         512           ReLU-21       [[1, 128, 80, 80]]    [1, 128, 80, 80]          0           Conv2D-28      [[1, 128, 80, 80]]    [1, 128, 80, 80]       147,584      BatchNorm-28     [[1, 128, 80, 80]]    [1, 128, 80, 80]         512           ReLU-22       [[1, 128, 80, 80]]    [1, 128, 80, 80]          0         conv_block-8     [[1, 256, 80, 80]]    [1, 128, 80, 80]          0          Upsample-4      [[1, 128, 80, 80]]   [1, 128, 160, 160]         0           Conv2D-29     [[1, 128, 160, 160]]  [1, 64, 160, 160]       73,792       BatchNorm-29    [[1, 64, 160, 160]]   [1, 64, 160, 160]         256           ReLU-23      [[1, 64, 160, 160]]   [1, 64, 160, 160]          0           up_conv-4      [[1, 128, 80, 80]]   [1, 64, 160, 160]          0           Conv2D-30     [[1, 64, 160, 160]]   [1, 32, 160, 160]        2,080       BatchNorm-30    [[1, 32, 160, 160]]   [1, 32, 160, 160]         128          Conv2D-31     [[1, 64, 160, 160]]   [1, 32, 160, 160]        2,080       BatchNorm-31    [[1, 32, 160, 160]]   [1, 32, 160, 160]         128           ReLU-24      [[1, 32, 160, 160]]   [1, 32, 160, 160]          0           Conv2D-32     [[1, 32, 160, 160]]    [1, 1, 160, 160]         33         BatchNorm-32     [[1, 1, 160, 160]]    [1, 1, 160, 160]          4           Sigmoid-4      [[1, 1, 160, 160]]    [1, 1, 160, 160]          0       Attention_block-4          []           [1, 64, 160, 160]          0           Conv2D-33     [[1, 128, 160, 160]]  [1, 64, 160, 160]       73,792       BatchNorm-33    [[1, 64, 160, 160]]   [1, 64, 160, 160]         256           ReLU-25      [[1, 64, 160, 160]]   [1, 64, 160, 160]          0           Conv2D-34     [[1, 64, 160, 160]]   [1, 64, 160, 160]       36,928       BatchNorm-34    [[1, 64, 160, 160]]   [1, 64, 160, 160]         256           ReLU-26      [[1, 64, 160, 160]]   [1, 64, 160, 160]          0         conv_block-9    [[1, 128, 160, 160]]  [1, 64, 160, 160]          0           Conv2D-35     [[1, 64, 160, 160]]    [1, 4, 160, 160]         260      =============================================================================Total params: 34,894,392Trainable params: 34,863,144Non-trainable params: 31,248-----------------------------------------------------------------------------Input size (MB): 0.29Forward/backward pass size (MB): 563.67Params size (MB): 133.11Estimated Total Size (MB): 697.07-----------------------------------------------------------------------------

{'total_params': 34894392, 'trainable_params': 34863144}

模型训练

train_dataset = PetDataset(mode='train') # 训练数据集val_dataset = PetDataset(mode='test') # 验证数据集optim = paddle.optimizer.RMSProp(learning_rate=0.001,                                  rho=0.9,                                  momentum=0.0,                                  epsilon=1e-07,                                  centered=False,                                 parameters=model.parameters())model.prepare(optim, paddle.nn.CrossEntropyLoss(axis=1))model.fit(train_dataset,           val_dataset,           epochs=15,           batch_size=32,          verbose=1)

模型预测

predict_dataset = PetDataset(mode='predict')predict_results = model.predict(predict_dataset)

Predict begin...step 1108/1108 [==============================] - 20ms/step         Predict samples: 1108

plt.figure(figsize=(10, 10))i = 0mask_idx = 0with open('./predict.txt', 'r') as f:    for line in f.readlines():        image_path, label_path = line.strip().split('\t')        resize_t = T.Compose([            T.Resize(IMAGE_SIZE)        ])        image = resize_t(PilImage.open(image_path))        label = resize_t(PilImage.open(label_path))        image = np.array(image).astype('uint8')        label = np.array(label).astype('uint8')        if i > 8:             break        plt.subplot(3, 3, i + 1)        plt.imshow(image)        plt.title('Input Image')        plt.axis("off")        plt.subplot(3, 3, i + 2)        plt.imshow(label, cmap='gray')        plt.title('Label')        plt.axis("off")                # 模型只有一个输出，通过predict_results[0]来取出1000个预测的结果        # 映射原始图片的index来取出预测结果，提取mask进行展示        data = predict_results[0][mask_idx][0].transpose((1, 2, 0))        mask = np.argmax(data, axis=-1)        plt.subplot(3, 3, i + 3)        plt.imshow(mask.astype('uint8'), cmap='gray')        plt.title('Predict')        plt.axis("off")        i += 3        mask_idx += 1plt.show()

你是一名 AI 行业编辑，请围绕下面这条热点输出一份资讯解读：
热点：基于Attention U-Net的宠物图像分割要求：
1. 先用一句话解释这条热点在讲什么
2. 再总结它为什么重要
3. 说明会影响哪些 AI 产品或内容方向
4. 最后给出 3 个适合资讯站使用的标题