{"id":6593,"date":"2022-08-30T10:01:00","date_gmt":"2022-08-30T02:01:00","guid":{"rendered":"http:\/\/139.9.1.231\/?p=6593"},"modified":"2022-08-29T10:01:58","modified_gmt":"2022-08-29T02:01:58","slug":"unet-2","status":"publish","type":"post","link":"http:\/\/139.9.1.231\/index.php\/2022\/08\/30\/unet-2\/","title":{"rendered":"Unet++"},"content":{"rendered":"\n

\u4ee3\u7801\uff1ahttps:\/\/github.com\/4uiiurz1\/pytorch-nested-unet\/blob\/master\/archs.py<\/a><\/p>\n\n\n\n

\u8bba\u6587\uff1ahttps:\/\/arxiv.org\/abs\/1807.10165<\/a><\/p>\n\n\n\n

\u8bba\u6587\u4f5c\u8005\u7684\u77e5\u4e4e\u8bb2\u89e3:https:\/\/zhuanlan.zhihu.com\/p\/44958351<\/a><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

UNet\u7684\u7f16\u89e3\u7801\u7ed3\u6784\u4e00\u7ecf\u63d0\u51fa\u4ee5\u6765\uff0c\u5927\u6709\u7edf\u4e00\u6df1\u5ea6\u5b66\u4e60\u56fe\u50cf\u5206\u5272\u4e4b\u52bf\uff0c\u540e\u7eed\u57fa\u4e8eUNet\u7684\u6539\u8fdb\u65b9\u6848\u4e5f\u7ecf\u4e45\u4e0d\u8870\uff0c\u4e00\u4e9b\u7814\u7a76\u8005\u4e5f\u5728\u4ece\u7f51\u7edc\u7ed3\u6784\u672c\u8eab\u6765\u601d\u8003UNet\u7684\u6709\u6548\u6027\u3002\u6bd4\u5982\u8bf4\u7f16\u89e3\u7801\u7f51\u7edc\u5e94\u8be5\u53d6\u51e0\u5c42\uff0c\u8df3\u8dc3\u8fde\u63a5\u662f\u5426\u80fd\u591f\u6709\u66f4\u591a\u7684\u53d8\u5316\u4ee5\u53ca\u4ec0\u4e48\u6837\u7684\u7ed3\u6784\u8bad\u7ec3\u8d77\u6765\u66f4\u52a0\u6709\u6548\u7b49\u95ee\u9898\u3002UNet\u672c\u8eab\u662f\u9488\u5bf9\u533b\u5b66\u56fe\u50cf\u5206\u5272\u4efb\u52a1\u800c\u63d0\u51fa\u6765\u7684\u7f51\u7edc\u7ed3\u6784\uff0c\u8be5\u4efb\u52a1\u4e0d\u50cf\u81ea\u7136\u56fe\u50cf\u5206\u5272\uff0c\u5bf9\u5206\u5272\u7cbe\u5ea6\u8981\u6c42\u5e76\u4e0d\u662f\u5341\u5206\u4e25\u683c\u3002\u4f46\u5bf9\u4e8e\u533b\u5b66\u56fe\u50cf\u800c\u8a00\uff0c\u5668\u5b98\u548c\u75c5\u7076\u7684\u5206\u5272\u5219\u8981\u6c42\u6781\u9ad8\u7684\u7cbe\u786e\u6027\uff0c\u56e0\u4e3a\u5f88\u591a\u65f6\u5019\u5206\u5272\u6548\u679c\u7684\u597d\u574f\u76f4\u63a5\u5173\u7cfb\u5230\u5bf9\u5e94\u7684\u4e34\u5e8a\u8bca\u65ad\u51b3\u7b56<\/strong>\u3002\u51fa\u4e8e\u4e0a\u8ff0\u4e24\u4e2a\u65b9\u9762\u7684\u52a8\u673a\uff0c\u5373\u8bbe\u8ba1\u66f4\u597d\u7684UNet\u7ed3\u6784\u548c\u63d0\u5347\u533b\u5b66\u56fe\u50cf\u5206\u5272\u7684\u7cbe\u5ea6\uff0c\u76f8\u5173\u7814\u7a76\u8005\u63d0\u51fa\u4e86\u4e00\u79cd\u5d4c\u5957\u7684UNet\u7ed3\u6784\uff08Nested UNet\uff09\uff0c\u4e5f\u53ebUNet++\uff0c\u63d0\u51faUNet++\u7684\u8bba\u6587\u4e3aUNet++: A Nested U-Net Architecture for Medical Image Segmentation\uff0c\u53d1\u8868\u4e8e2018\u5e74\u7684\u533b\u5b66\u56fe\u50cf\u8ba1\u7b97\u548c\u8ba1\u7b97\u673a\u8f85\u52a9\u5e72\u9884\uff08Medical Image Computing and Computer Assisted Intervention\uff0cMICCAI\uff09\u4f1a\u8bae\u4e0a\u3002<\/p>\n\n\n\n

UNet++\u53d6\u540d\u4e3a\u5d4c\u5957\u7684UNet\uff0c\u5c31\u5728\u4e8e\u5176\u6574\u4f53\u7f16\u89e3\u7801\u7f51\u7edc\u7ed3\u6784\u4e2d\u8fd8\u5d4c\u5957\u4e86\u7f16\u89e3\u7801\u7684\u5b50\u7f51\u7edc\uff08sub-networks\uff09\uff0c\u5728\u6b64\u57fa\u7840\u4e0a\u91cd\u65b0\u8bbe\u8ba1UNet\u4e2d\u95f4\u7684\u8df3\u8dc3\u8fde\u63a5\uff0c\u5e76\u8865\u5145\u4e86\u6df1\u76d1\u7763\u673a\u5236\u52a0\u901f\u7f51\u7edc\u8bad\u7ec3\u6536\u655b\u3002\u5b8c\u6574\u7684UNet++\u7ed3\u6784\u5982\u4e0b\u56fe\u6240\u793a\u3002<\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

\u56fe\u4e2d\u9ed1\u8272\u90e8\u5206\u4e3a\u539f\u59cb\u7684UNet\u7ed3\u6784\uff0c\u5305\u62ec\u7f16\u7801\u5668\u4e0b\u91c7\u6837\u3001\u89e3\u7801\u5668\u4e0a\u91c7\u6837\u548c\u9ed1\u8272\u865a\u7ebf\u7684\u8df3\u8dc3\u8fde\u63a5\u4e09\u4e2a\u90e8\u5206\uff1b\u7eff\u8272\u90e8\u5206\u5373\u5d4c\u5957\u7684UNet\u5b50\u7f51\u7edc\uff0c\u5305\u62ec\u5377\u79ef\u548c\u4e0a\u91c7\u6837\u4e24\u90e8\u5206\uff0c\u800c\u84dd\u8272\u865a\u7ebf\u90e8\u5206\u5c31\u662fUNet++\u91cd\u65b0\u8bbe\u8ba1\u540e\u7684\u8df3\u8dc3\u8fde\u63a5\uff0c\u8fd9\u90e8\u5206\u8ddfDenseNet\u7684\u5bc6\u96c6\u8fde\u63a5\u7c7b\u4f3c\uff0c\u8fd9\u91cc\u662f\u4e3a\u5b50\u7f51\u7edc\u63d0\u4f9b\u8df3\u8dc3\u8fde\u63a5\uff1b\u6700\u4e0a\u9762\u7ea2\u9ed1\u8fde\u7ebf\u5219\u662fUNet++\u8865\u5145\u7684\u6df1\u76d1\u7763\u673a\u5236\uff0c\u76ee\u7684\u662f\u4e3a\u4e86\u7f51\u7edc\u80fd\u591f\u987a\u5229\u5f97\u5230\u8bad\u7ec3\u3002<\/p>\n\n\n\n

\u4e0b\u9762\u6211\u4eec\u4ece\u7ed3\u6784\u8bbe\u8ba1\u7684\u89d2\u5ea6\u6765\u5bf9UNet++\u8fdb\u884c\u89e3\u8bfb\u3002\u5173\u4e8eUNet\u7ed3\u6784\uff0c\u6700\u9996\u8981\u7684\u95ee\u9898\u5c31\u662f\u7f51\u7edc\u5e94\u8be5\u6709\u51e0\u5c42\uff0c\u539f\u59cb\u7684UNet\u7ed3\u6784\u7528\u4e864\u5c42\u4e0b\u91c7\u6837\u548c4\u5c42\u4e0a\u91c7\u6837\uff0c\u90a3\u4e48\u662f\u4e0d\u662f4\u5c42\u5c31\u8db3\u4ee5\u6ee1\u8db3\u6240\u6709\u7684\u5206\u5272\u4efb\u52a1\u9700\u8981\uff1f\u7b54\u6848\u662f\u5426\u5b9a\u7684\u3002\u901a\u8fc7\u672c\u8282\u4e4b\u524d\u7684\u7f51\u7edc\u7ed3\u6784\u5206\u6790\uff0c\u6211\u4eec\u5df2\u7ecf\u77e5\u9053\uff0c\u6d45\u5c42\u7f51\u7edc\u80fd\u591f\u63d0\u53d6\u56fe\u50cf\u7c97\u7c92\u5ea6\u7279\u5f81\uff0c\u83b7\u53d6\u56fe\u50cf\u57fa\u672c\u5f62\u6001\uff1b\u6df1\u5c42\u7f51\u7edc\u80fd\u591f\u63d0\u53d6\u56fe\u50cf\u7684\u62bd\u8c61\u7279\u5f81\uff0c\u83b7\u53d6\u56fe\u50cf\u8bed\u4e49\u4fe1\u606f\uff0c\u603b\u4e4b\u6d45\u6709\u6d45\u7684\u4fa7\u91cd\uff0c\u6df1\u6709\u6df1\u7684\u597d\u5904\u3002\u540c\u4e4b\u524dRefineNet\u7684\u89c2\u70b9\u4e00\u6837\uff0cUNet++\u7684\u4f5c\u8005\u8ba4\u4e3a\uff0c\u4e0d\u7ba1\u662f\u6d45\u5c42\u3001\u6df1\u5c42\u8fd8\u662f\u4e2d\u5c42\uff0c\u6240\u6709\u5c42\u6b21\u7684\u7279\u5f81\u5bf9\u4e8e\u6700\u540e\u7684\u5206\u5272\u90fd\u662f\u91cd\u8981\u7684\u3002\u6709\u7684\u6570\u636e\u5206\u5272\u4efb\u52a1\u7b80\u5355\uff0c\u56fe\u50cf\u4fe1\u606f\u5355\u4e00\uff0c\u53ef\u80fd\u6d45\u5c42\u7f51\u7edc\u5c31\u8db3\u4ee5\u8fbe\u5230\u5f88\u597d\u7684\u6548\u679c\uff0c\u800c\u6709\u7684\u6570\u636e\u4efb\u52a1\u590d\u6742\uff0c\u56fe\u50cf\u4fe1\u606f\u4e30\u5bcc\uff0c\u53ef\u80fd\u9700\u8981\u66f4\u6df1\u5c42\u7684\u7f51\u7edc\u7ed3\u6784\u624d\u80fd\u8fbe\u5230\u4e0d\u9519\u7684\u6548\u679c\uff0c\u4e4b\u524d\u7684UNet\u7ed3\u6784\u8bbe\u8ba1\u5f88\u96be\u540c\u65f6\u7167\u987e\u5230\u8fd9\u79cd\u666e\u9002\u6027\u3002\u800cUNet++\u901a\u8fc7\u8bbe\u8ba1\u4e0d\u540c\u6df1\u5ea6\u7684\u5d4c\u5957UNet\u5b50\u7f51\u7edc\u6765\u5b9e\u73b0\u8fd9\u79cd\u666e\u9002\u6027\uff0c\u6240\u4ee5UNet\u7684\u6df1\u5ea6\u5230\u8fd9\u91cc\u5c31\u89e3\u51b3\u4e86\u3002<\/p>\n\n\n\n

\u7b2c\u4e8c\u4e2a\u95ee\u9898\u5219\u662f\u52a0\u5165\u4e0d\u540c\u6df1\u5ea6\u7684\u5d4c\u5957\u7f51\u7edc\u540e\uff0c\u8df3\u8dc3\u8fde\u63a5\u90e8\u5206\u8be5\u5982\u4f55\u8c03\u6574\u3002\u5728UNet\u4e2d\uff0c\u8df3\u8dc3\u8fde\u63a5\u7531\u540c\u5c42\u7f16\u7801\u5668\u76f4\u8fde\u5230\u7f16\u7801\u5668\u4e0a\u91c7\u6837\u5bf9\u5e94\u5c42\u3002\u4f46\u52a0\u5165\u5d4c\u5957\u5b50\u7f51\u7edc\u540e\uff0cUNet\u4e2d\u539f\u5148\u7684\u957f\u8fde\u63a5\u5c31\u4e0d\u590d\u5b58\u5728\u4e86\uff0c\u53d6\u800c\u4ee3\u4e4b\u7684\u662f\u5404\u5b50\u7f51\u7edc\u4e2d\u7684\u77ed\u8fde\u63a5\u3002UNet++\u7684\u4f5c\u8005\u4eec\u8ba4\u4e3a\uff0c\u957f\u8fde\u63a5\u5728UNet\u4e2d\u662f\u6709\u5fc5\u8981\u7684\uff0c\u80fd\u591f\u5c06\u56fe\u50cf\u4e2d\u524d\u540e\u4fe1\u606f\u8054\u7cfb\u8d77\u6765\uff0c\u5bf9\u4e8e\u4e0b\u91c7\u6837\u9020\u6210\u7684\u4fe1\u606f\u635f\u5931\u6709\u5f88\u597d\u7684\u8865\u5145\u4f5c\u7528\u3002\u6240\u4ee5\uff0cUNet++\u53c8\u53c2\u8003DenseNet\u7684\u5bc6\u96c6\u8fde\u63a5\u8bbe\u8ba1\uff0c\u7ed9\u5d4c\u5957\u7f51\u7edc\u8865\u5145\u4e86\u957f\u8fde\u63a5\uff0c\u5982\u4e0a\u56fe\u6240\u793a\u3002<\/p>\n\n\n\n

\u4f46\u662f\u8fd9\u6837\u53c8\u5e26\u6765\u4e86\u7b2c\u4e09\u4e2a\u95ee\u9898\uff1a\u53cd\u5411\u4f20\u64ad\u7684\u65f6\u5019\u4e2d\u95f4\u90e8\u5206\u53ef\u80fd\u4f1a\u6536\u4e0d\u5230\u7531\u635f\u5931\u51fd\u6570\u53cd\u4f20\u56de\u6765\u7684\u68af\u5ea6\u3002\u6240\u4ee5\u89c1\u62db\u62c6\u62db\uff0cUNet++\u53c8\u901a\u8fc7\u6df1\u76d1\u7763\u7684\u65b9\u6cd5\u6765\u5f3a\u884c\u52a0\u68af\u5ea6\uff0c\u5e2e\u52a9\u7f51\u7edc\u6b63\u5e38\u8fdb\u884c\u8bad\u7ec3\u3002\u4f46\u6df1\u76d1\u7763\u5bf9\u4e8eUNet++\u7684\u597d\u5904\u7edd\u4e0d\u4ec5\u4ec5\u9650\u4e8e\u6b64\uff0c\u901a\u8fc7\u4e0d\u540c\u6df1\u76d1\u7763\u635f\u5931\u51fd\u6570\uff0cUNet++\u53ef\u4ee5\u901a\u8fc7\u7f51\u7edc\u526a\u679d\u6765\u5b9e\u73b0\u53ef\u4f38\u7f29\u6027\u3002\u6240\u4ee5\uff0c\u603b\u7ed3\u6765\u8bf4UNet++\u76f8\u8f83\u4e8e\u539f\u59cb\u7684UNet\uff0c\u6709\u5982\u4e0b\u4e24\u4e2a\u4f18\u52bf\uff1a<\/p>\n\n\n\n

\uff081\uff09\u901a\u8fc7\u5d4c\u5957\u5b50\u7f51\u7edc\u548c\u957f\u77ed\u8fde\u63a5\u6765\u6574\u5408\u4e0d\u540c\u5c42\u6b21\u7684\u56fe\u50cf\u7279\u5f81\uff0c\u4f7f\u5f97\u7f51\u7edc\u5206\u5272\u7cbe\u5ea6\u66f4\u9ad8\uff1b<\/p>\n\n\n\n

\uff082\uff09\u7075\u6d3b\u7684\u7f51\u7edc\u7ed3\u6784\u914d\u5408\u6df1\u76d1\u7763\u673a\u5236\uff0c\u8ba9\u53c2\u6570\u91cf\u5de8\u5927\u7684\u6df1\u5ea6\u7f51\u7edc\u5728\u53ef\u63a5\u53d7\u7684\u7cbe\u5ea6\u8303\u56f4\u5185\u80fd\u591f\u5927\u5e45\u5ea6\u7684\u7f29\u51cf\u53c2\u6570\u91cf\u3002<\/p>\n\n\n\n

UNet++\u4e0eUNet\u7b49\u7f51\u7edc\u5206\u5272\u6548\u679c\u5bf9\u6bd4\u5982\u4e0b\u56fe\u6240\u793a\u3002<\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

UNet++\u4e5f\u8fdb\u4e00\u6b65\u58ee\u5927\u4e86UNet\u5bb6\u65cf\u7f51\u7edc\uff0c\u540e\u7eed\u57fa\u4e8e\u5176\u7684\u6539\u8fdb\u7248\u672c\u4e5f\u6709\u5f88\u591a\uff0c\u6bd4\u5982Attention UNet++\u3001UNet 3+\u7b49\u3002\u4e0b\u8ff0\u4ee3\u7801\u7ed9\u51fa\u4e86UNet++\u7684\u4e00\u4e2a\u5b9e\u73b0\u53c2\u8003\u3002<\/p>\n\n\n\n

\r\nclass NestedUNet(nn.Module):\r\n    def __init__(self, num_classes, input_channels=3, deep_supervision=False, **kwargs):\r\n        super().__init__()\r\n        nb_filter = [32, 64, 128, 256, 512]\r\n        self.deep_supervision = deep_supervision\r\n        self.pool = nn.MaxPool2d(2, 2)\r\n        self.up = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)\r\n\r\n        self.conv0_0 = VGGBlock(input_channels, nb_filter[0], nb_filter[0])\r\n        self.conv1_0 = VGGBlock(nb_filter[0], nb_filter[1], nb_filter[1])\r\n        self.conv2_0 = VGGBlock(nb_filter[1], nb_filter[2], nb_filter[2])\r\n        self.conv3_0 = VGGBlock(nb_filter[2], nb_filter[3], nb_filter[3])\r\n        self.conv4_0 = VGGBlock(nb_filter[3], nb_filter[4], nb_filter[4])\r\n\r\n        self.conv0_1 = VGGBlock(nb_filter[0]+nb_filter[1], nb_filter[0], nb_filter[0])\r\n        self.conv1_1 = VGGBlock(nb_filter[1]+nb_filter[2], nb_filter[1], nb_filter[1])\r\n        self.conv2_1 = VGGBlock(nb_filter[2]+nb_filter[3], nb_filter[2], nb_filter[2])\r\n        self.conv3_1 = VGGBlock(nb_filter[3]+nb_filter[4], nb_filter[3], nb_filter[3])\r\n\r\n        self.conv0_2 = VGGBlock(nb_filter[0]*2+nb_filter[1], nb_filter[0], nb_filter[0])\r\n        self.conv1_2 = VGGBlock(nb_filter[1]*2+nb_filter[2], nb_filter[1], nb_filter[1])\r\n        self.conv2_2 = VGGBlock(nb_filter[2]*2+nb_filter[3], nb_filter[2], nb_filter[2])\r\n\r\n        self.conv0_3 = VGGBlock(nb_filter[0]*3+nb_filter[1], nb_filter[0], nb_filter[0])\r\n        self.conv1_3 = VGGBlock(nb_filter[1]*3+nb_filter[2], nb_filter[1], nb_filter[1])\r\n        self.conv0_4 = VGGBlock(nb_filter[0]*4+nb_filter[1], nb_filter[0], nb_filter[0])\r\n\r\n        if self.deep_supervision:\r\n            self.final1 = nn.Conv2d(nb_filter[0], num_classes, kernel_size=1)\r\n            self.final2 = nn.Conv2d(nb_filter[0], num_classes, kernel_size=1)\r\n            self.final3 = nn.Conv2d(nb_filter[0], num_classes, kernel_size=1)\r\n            self.final4 = nn.Conv2d(nb_filter[0], num_classes, kernel_size=1)\r\n        else:\r\n            self.final = nn.Conv2d(nb_filter[0], num_classes, kernel_size=1)\r\n\r\n\r\n    def forward(self, input):\r\n        x0_0 = self.conv0_0(input)\r\n        x1_0 = self.conv1_0(self.pool(x0_0))\r\n        x0_1 = self.conv0_1(torch.cat([x0_0, self.up(x1_0)], 1))\r\n\r\n        x2_0 = self.conv2_0(self.pool(x1_0))\r\n        x1_1 = self.conv1_1(torch.cat([x1_0, self.up(x2_0)], 1))\r\n        x0_2 = self.conv0_2(torch.cat([x0_0, x0_1, self.up(x1_1)], 1))\r\n\r\n        x3_0 = self.conv3_0(self.pool(x2_0))\r\n        x2_1 = self.conv2_1(torch.cat([x2_0, self.up(x3_0)], 1))\r\n        x1_2 = self.conv1_2(torch.cat([x1_0, x1_1, self.up(x2_1)], 1))\r\n        x0_3 = self.conv0_3(torch.cat([x0_0, x0_1, x0_2, self.up(x1_2)], 1))\r\n\r\n        x4_0 = self.conv4_0(self.pool(x3_0))\r\n        x3_1 = self.conv3_1(torch.cat([x3_0, self.up(x4_0)], 1))\r\n        x2_2 = self.conv2_2(torch.cat([x2_0, x2_1, self.up(x3_1)], 1))\r\n        x1_3 = self.conv1_3(torch.cat([x1_0, x1_1, x1_2, self.up(x2_2)], 1))\r\n        x0_4 = self.conv0_4(torch.cat([x0_0, x0_1, x0_2, x0_3, self.up(x1_3)], 1))\r\n\r\n        if self.deep_supervision:\r\n            output1 = self.final1(x0_1)\r\n            output2 = self.final2(x0_2)\r\n            output3 = self.final3(x0_3)\r\n            output4 = self.final4(x0_4)\r\n            return [output1, output2, output3, output4]\r\n        else:\r\n            output = self.final(x0_4)\r\n            return output<\/code><\/pre>\n\n\n\n
\u6700\u540e\uff0c\u7ed9\u51fa\u4f5c\u8005\u5728\u77e5\u4e4e\u4e0a\u7684\u4e00\u6bb5\u603b\u7ed3\uff1a<\/p>\n\n\n\n
     \u56de\u987e\u4e00\u4e0b\u8fd9\u6b21\u5206\u4eab\uff0c\u6211\u95ee\u4e86\u597d\u591a\u95ee\u9898\uff0c\u4e5f\u63d0\u4f9b\u4e86\u5176\u4e2d\u4e00\u4e9b\u6211\u4e2a\u4eba\u7ed9\u51fa\u7684\u89e3\u91ca\u3002<\/p>\n\n\n\n
\u4e00\u987f\u542c\u4e0b\u6765\uff0c\u70ed\u5fc3\u7684\u7f51\u53cb\u53ef\u80fd\u4f1a\u61f5\u4e86\uff0c\u5c31\u8fd9\u4e48\u4e2a\u5206\u5272\u7f51\u7edc\uff0c\u90fd\u80fd\u8bf4\u8fd9\u4e48\u4e45\uff0c\u8981\u6211\u8bf4\u5c31\u653e\u4e2a\u7ed3\u6784\u56fe\uff0c\u8bf4\u8fd9\u4e2a\u7f51\u7edc\u5f88\u725b\u903c\uff0c\u518d\u544a\u77e5\u4e00\u4e0b\u4ee3\u7801\u5728\u54ea\u513f\uff0c\u8c22\u8c22\u5927\u5bb6\u5c31\u5b8c\u4e8b\u513f\u4e86\u3002<\/p>\n\n\n\n
    \u5176\u5b9e\u641e\u5b66\u672f\u505a\u7814\u7a76\u4e0d\u662f\u8fd9\u6837\u5b50\u7684\uff0cUNet++\u80af\u5b9a\u9a6c\u4e0a\u5c31\u4f1a\u88ab\u66f4\u5f3a\u7684\u7ed3\u6784\u6240\u4ee3\u66ff\uff0c\u4f46\u662f\u8981\u8bbe\u8ba1\u51fa\u66f4\u5f3a\u7684\u7ed3\u6784\uff0c\u4f60\u5f97\u9996\u5148\u660e\u767d\u8fd9\u4e2a\u7ed3\u6784\uff0c\u751a\u81f3\u5b83\u7684\u539f\u578bU-Net\u8bbe\u8ba1\u80cc\u540e\u7684\u5fc3\u8def\u5386\u7a0b\u3002\u4e0e\u5176\u548c\u5927\u5bb6\u5206\u4eab\u4e00\u4e2a\u82cd\u767d\u7684\u5206\u5272\u7f51\u7edc\uff0c\u6211\u66f4\u613f\u610f\u5206\u4eab\u7684\u662f\u8fd9\u4e2a\u9879\u76ee\u80cc\u540e\u4ece\u5f00\u59cb\u8ba4\u8bc6U-Net\uff0c\u5230\u5206\u6790\u5b83\u7684\u7ec4\u6210\uff0c\u5230\u6279\u5224\u6027\u7684\u89e3\u8bfb\uff0c\u518d\u5230\u6539\u8fdb\u601d\u8def\u7684\u5f62\u6210\uff0c\u5b9e\u9a8c\u8bbe\u8ba1\uff0c\u50cf\u521a\u521a\u5206\u4eab\u8fc7\u7a0b\u4e2d\u4e00\u6b21\u6b21\u5c34\u5c2c\u7684\u81ea\u95ee\u81ea\u7b54\uff0c\u4e2d\u95f4\u90a3\u4e9b\u975e\u5e38\u9971\u6ee1\u7684\u5fc3\u8def\u5386\u7a0b\u3002\u8fd9\u4e5f\u662f\u6211\u5728\u535a\u58eb\u7684\u4e24\u5e74\u4e2d\u5b66\u5230\u7684\u505a\u7814\u7a76\u7684\u8303\u5f0f\u3002<\/p>\n\n\n\n
\u8bf4\u53e5\u9898\u5916\u8bdd\uff0c\u5c31\u8ddf\u73a9\u72fc\u4eba\u6740\u4e00\u6837\uff0c\u4f60\u4e00\u4e0a\u6765\u5c31\u8bf4\u81ea\u5df1\u9884\u8a00\u5bb6\uff0c\u9a8c\u4e86\u8c01\u8c01\u8c01\uff0c\u90a3\u6ca1\u4eba\u4fe1\u7684\u5440\uff0c\u4f60\u5f97\u8bf4\u6e05\u695a\u4e3a\u4ec0\u4e48\u8981\u9a8c\u4ed6\uff0c\u8b66\u5fbd\u600e\u4e48\u7559\uff0c\u628a\u8fd9\u4e9b\u5fc3\u8def\u5386\u7a0b\u90fd\u76d8\u76d8\u6e05\u695a\uff0c\u8eab\u4efd\u624d\u80fd\u505a\u5b9e\u3002<\/p>\n\n\n\n
\u6211\u5728\u5fae\u535a\u4e0a\u4e5f\u770b\u5230\u6709\u4eba\u8bf4\uff0c\u4e5f\u60f3\u5230\u4e86\u975e\u5e38\u7c7b\u4f3c\u7684\u7ed3\u6784\uff0c\u5b9e\u9a8c\u4e5f\u5feb\u505a\u5b8c\u4e86\uff0c\u770b\u5230\u4e86\u6211\u8fd9\u7bc7\u8bba\u6587\u5fc3\u5c31\u51c9\u4e86\u3002\u5176\u5b9e\u7f51\u7edc\u7ed3\u6784\u600e\u4e48\u6837\u771f\u7684\u4e0d\u91cd\u8981\uff0c\u91cd\u8981\u7684\u662f\u4f60\u600e\u4e48\u80fd\u628a\u6545\u4e8b\u7ed9\u8bb2\u6e05\u695a\uff0c\u8981\u662f\u8bb2\u5b8c\u4ee5\u540e\u8fd8\u80fd\u591f\u5f15\u8d77\u66f4\u591a\u7684\u601d\u8003\u548c\u8ba8\u8bba\u90a3\u5c31\u66f4\u597d\u4e86\u3002\u6211\u5728\u5206\u4eab\u4e2d\u63d0\u5230\u4e86\u5f88\u591a\u6211\u4eec\u8bba\u6587\u4e2d\u7684\u4e0d\u8db3\u4e4b\u5904\uff0c\u4e5f\u975e\u5e38\u6b22\u8fce\u5927\u5bb6\u53ef\u4ee5\u6279\u8bc4\u6307\u6b63\u3002<\/p>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
\u4ee3\u7801\uff1ahttps:\/\/github.com\/4uiiurz1\/pytorch-nested-unet\/blob … \u7ee7\u7eed\u9605\u8bfbUnet++<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[24],"tags":[],"_links":{"self":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/6593"}],"collection":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/comments?post=6593"}],"version-history":[{"count":15,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/6593\/revisions"}],"predecessor-version":[{"id":6611,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/posts\/6593\/revisions\/6611"}],"wp:attachment":[{"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/media?parent=6593"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/categories?post=6593"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/139.9.1.231\/index.php\/wp-json\/wp\/v2\/tags?post=6593"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}