explore.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 第四章 图像分割"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.1 什么是图像分割？\n",
    "\n",
    "　　图像分割就是给定一张图片，预测图像中每个像素的类别。它有三个子问题：一是只输出目标类别，称为语义分割（如图１(c)）；二是区分不同目标的实例，称为实例分割（如图１(d)）；三是区分图像中所有的物体，包括天空，道路，建筑，沙滩，草地等场景，称为全景分割（如图２中的(d)）。\n",
    "  \n",
    "<center><img src=\"output/book01.jpg\" />图１ 图像识别的核心任务</center><br>\n",
    "<center><img src=\"output/book02.jpg\" />图2 全景分割实例</center>\n",
    "\n",
    "　　语义分割和实例分割要分割的目标，通常具有一定的边界，如人，车等常规目标，是可数的，全景分割中包含不可数目标，所以，从范围上说，全景分割包含语义分割和实例分割，但全景分割不允许同一像素归为不同的目标，它带来的好处是可以在图上进行像素级别的推理。显然，全景分割难度最高，应用价值最大，实例分割次之，语义分割最简单，现在主流分割算法主要解决实例分割，本章也集中精力研究实例分割算法Mask R-CNN。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.2 常用数据集\n",
    "　　主要介绍四个：Pascal VOC，CityScapes，MS COCO，ADE20K。\n",
    "  \n",
    "　　第一个是Pascal VOC数据集，这是一个老牌数据集，它提供20个类别，包括人，车等。2012版有9993张标注图，提供了类别标注和实例标注，既可以做语义分割，也可以做实例分割，把不同的人车标记出来。\n",
    "  \n",
    "<center><img src=\"output/book03.jpg\" />图3 Pascal VOC分割例子</center>\n",
    "\n",
    "　　第二个是CityScapes数据集，主要面向道路驾驶场景，有30个类，图片来自50城市，其中有5000张图进行了精细标注，精确到像素级别， 20000 张图片有粗糙的标注，可以做语义分割和实例分割。\n",
    "  \n",
    "<center><img src=\"output/book04.jpg\" />图4 CityScapes精标图</center><br>\n",
    "<center><img src=\"output/book05.jpg\" />图5 CityScapes粗标图</center>\n",
    "\n",
    "　　第三个是MS COCO数据集，这是目前语义分割的最大数据集，80类，超过30万张图片，其中20万张有标注，实例超过200万个，最新论文一般都会在MS COCO数据集上做实验，因为它难度最大，挑战性最高。图１就是来自于MS COCO的样本，本章后面结合Mask R-CNN有更详细的分析。\n",
    "    \n",
    "　　第四个是ADE20K数据集，训练集含20210张图片，验证集含2000张图片，每张图像，目标和部件分割被存储为不同的png文件，分割从上到下，包含天空、墙、建筑和地板等场景。\n",
    "<center><img src=\"output/book06.jpg\" />图6 ADE20K标注实例</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.3 传统分割算法\n",
    "\n",
    "　　传统图像分割最具代表性的就是图切割算法GraphCut和GrabCut，它们都需要提供前景和背景像素种子，通过种子建立前景背景概率分布模型来估算所有像素属于前景和背景的概率。直观上，种子越多越精确，分割精度就越高，当种子无法覆盖所有分布时，会影响分割的准确度。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.3.1 GraphCut\n",
    "　　将图像中的每个像素看成图中的一个节点，首先在图中增加两个节点，分别为F和B，F代表前景，而B代表背景，其次每两个相邻像素点用一条边连起来，每个像素点和F点用一条边连起来，每个像素点和B点也用一条边连起来，形成一个带权重的无向图，如下图：\n",
    "  \n",
    "<center><img src=\"output/book07.jpg\" />图7 GraphCut无向图</center>\n",
    " \n",
    "　　一个切割就是将图像分成两部分，第一部分的像素点和F相连为前景，第二部分的像素点和B相连为背景，切割通过优化以下能量函数来实现：\n",
    "\n",
    "<center><img src=\"output/book08.jpg\" />图8 GraphCut能量函数</center>\n",
    "  \n",
    "　　其中g(X)为数据(Data)项，由像素v属于前景和属于背景的概率决定，h(v,u)称为平滑(Smoothness)项，由像素v和其邻接像素u的相似程度决定。最简单的前景背景概率计算就是基于图像的直方图，而衡量两个相邻像素点相似度最简单的方法则是用灰度差值。通过上式可以看到，能量的最优化就是将最有可能是前景的列为一组，最有可能是背景的列为另一组，然后将灰度差最大的相邻像素切割开来，这样的分割是全局最优分割。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.3.2 GrabCut\n",
    "\n",
    "　　GrabCut可看成GraphCut的升级版，主要改进是：一是简化种子的选取方法，用户只要用一个矩形框住前景；二是迭代优化前景和背景分布的概率模型ＧＭＭ－－高斯混合模型。具体步骤如下：\n",
    "   \n",
    "　　1. 用矩形外的像素作为背景，矩形内的像素作为前景，分别训练背景和前景GMM；\n",
    "  \n",
    "　　2. 用训练好的GMM来计算每个像素属于背景和前景的概率，即计算出能量函数E中的Data项，计算Smoothness项的方法与GraphCut大致相同；\n",
    "  \n",
    "　　3. 用最优化能量函数完成图的最小分割，从而完成图像分割；\n",
    "  \n",
    "　　4. 用3中的分割结果（前景和背景像素）再次训练前景和背景GMM；\n",
    "  \n",
    "　　5. 重复步骤2，3，4，直到分割结果收敛。\n",
    "  \n",
    "　　从上述步骤可以看出，GrabCut是一个迭代算法，目标是分割能量最小，期望最大化（EM--Expectation Maximization）。它假定用户提供的矩形内部是前景，外部是背景，遗憾地是，内部也可能有背景像素，所以种子是不完全正确的，更严重的是存在仅凭颜色无法区分背景和前景的系统性难题。好在GMM模型并不要求所有初始数据都正确，即使有部分不正确，也可以通过EM步骤得到适当纠正，GrabCut正是利用了GMM的这一特性才表现良好。值得注意的是，GMM有可能陷入局部最优，所以GrabCut也有此问题。\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 实战\n",
    "\n",
    "　　OpenCV实现了GrabCut算法，下面代码是一个简易包装，可完成抠图，程序首先对图片缩放，然后选取种子矩形(距图片边界10个像素)，最后迭代16次输出。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVwAAAD7CAYAAAAmeCzOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6dd828a7f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVwAAAD7CAYAAAAmeCzOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngStiAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6db161a320>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 7.75 s, sys: 222 ms, total: 7.97 s\n",
      "Wall time: 12.3 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# tools/grabcut.py\n",
    "import sys\n",
    "import cv2\n",
    "import random\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "RESIZE_WIDTH = 320\n",
    "RESIZE_HEIGHT = 480\n",
    "\n",
    "\n",
    "def resize_grabcut(img):\n",
    "    \"\"\"Image Resize and Grab Cut.\"\"\"\n",
    "    img = cv2.resize(img, (RESIZE_WIDTH, RESIZE_HEIGHT))\n",
    "\n",
    "    mask = np.zeros((img.shape[:2]), np.uint8)\n",
    "    bgmodel = np.zeros((1, 65), np.float64)\n",
    "    fgmodel = np.zeros((1, 65), np.float64)\n",
    "\n",
    "    border = 10\n",
    "    rect = (border, border, img.shape[1] - border, img.shape[0] - border)\n",
    "\n",
    "    cv2.grabCut(img, mask, rect, bgmodel, fgmodel, 16, cv2.GC_INIT_WITH_RECT)\n",
    "\n",
    "    # 0 -- cv2.GC_BGD, 1 -- cv2.GC_FGD, 2 -- cv2.GC_PR_BGD, 3 -- cv2.GC_PR_FGD\n",
    "    mask2 = np.where((mask == 2) | (mask == 0), 0, 1).astype('uint8')\n",
    "\n",
    "    return img * mask2[:, :, np.newaxis]\n",
    "\n",
    "\n",
    "def images_show(image1, image2, title1=None, title2=None):\n",
    "    \"\"\"Show images.\"\"\"\n",
    "    plt.subplot(1, 2, 1)\n",
    "    plt.title('image 1' if not title1 else title1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.imshow(cv2.cvtColor(image1, cv2.COLOR_BGR2RGB))\n",
    "\n",
    "    plt.subplot(1, 2, 2)\n",
    "    plt.title('image 2' if not title2 else title2)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.imshow(cv2.cvtColor(image2, cv2.COLOR_BGR2RGB))\n",
    "\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "def grabcut(image_filename):\n",
    "    \"\"\"Two stages image grab cut.\"\"\"\n",
    "    image = cv2.imread(image_filename)\n",
    "    image = cv2.resize(image, (RESIZE_WIDTH, RESIZE_HEIGHT))\n",
    "\n",
    "    result = resize_grabcut(image)\n",
    "\n",
    "    images_show(image, result, 'Orignal', 'GrabCut')\n",
    "\n",
    "\n",
    "grabcut(\"images/messi.jpg\")\n",
    "grabcut(\"images/girl.jpg\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 改进方向\n",
    "　　可以看到，GrabCut运行慢，抠图效果一般。效果方面，有两个改进方向：一是毛边严重，可采用导向滤波等算法改进；二是抠图错误，可采用深度学习方法改进，这里有一个连接供您参考：\n",
    "https://github.com/jfzhang95/DeepGrabCut-PyTorch.git\n",
    "。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.4 Mask R-CNN\n",
    "\n",
    "　　传统图像分割算法只利用了图像本身的颜色、纹理等底层信息，无法从高层语义信息中获取增强线索，所以正确的图像分割算法必然会借助深度学习的方法来逾越底层到高层的信息鸿沟，Mask R-CNN就是一个极其优秀的例子，它由何凯明博士团队发明，荣获ICCV2017最佳论文奖（Marr Prize），论文地址：https://arxiv.org/abs/1703.06870 。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.4.1 概述\n",
    "> #### 摘要\n",
    ">\n",
    ">　　我们提出了一个概念上简单，灵活和通用的目标分割框架。我们的方法能有效地检测图像中的目标，同时为每个实例生成高质量的分割掩码。称为Mask R-CNN的方法通过添加一个与现有目标检测框回归并行的，用于预测目标掩码的分支来扩展Faster R-CNN。Mask R-CNN训练简单，相对于Faster R-CNN，只需增加一个较小的开销，运行速度可达5 FPS。此外，Mask R-CNN很容易推广到其他任务，例如，允许我们在同一个框架中估计人的姿势。我们在COCO挑战的所有三个项目中取得了最佳成绩，包括目标分割，目标检测和人体关键点检测。在没有使用额外技巧的情况下，Mask R-CNN优于所有现有的单一模型，包括COCO 2016挑战优胜者。我们希望我们的简单而有效的方法将成为一个促进未来目标级识别领域研究的坚实基础。我们稍后将提供代码。\n",
    "\n",
    ">    来自论文《Mask R-CNN》的中文翻译\n",
    "\n",
    "　　**Mask R-CNN是什么，可以做哪些任务？**\n",
    "\n",
    "　　简单地说，Mask R-CNN就是一个通用的实例分割架构，可以用来做“目标检测”、“实例分割”、“关键点检测”等任务。\n",
    "\n",
    "　　下面是三个例子，很明显，它几乎可以直接应用到“自动驾驶”，\"人体姿态检测\"，“智慧医疗”等领域。\n",
    "\n",
    "<center><img src=\"output/book09.jpg\" />图9 分割实例</center><br>\n",
    "<center><img src=\"output/book10.jpg\" />图10 人体关键点检测</center><br>\n",
    "<center><img src=\"output/book11.jpg\" />图11 分割细胞核</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.4.2 架构\n",
    "  \n",
    "　　**Mask R-CNN的架构长什么样？创新在哪里?**\n",
    "   \n",
    "　　论文《Mask R-CNN》明确告诉我们：Mask R-CNN就是Faster R-CNN的直接扩展，Faster R-CNN已有目标分类和目标检测模块，Mask R-CNN扩展的就是目标分割模块，整体模型如图12所示。这里，Mask是掩码的英文名，代表分割，CNN是Convolutional Neural Network的缩写，中文名叫卷积神经网，R是Region的缩写，表示在区域上工作。\n",
    "  \n",
    "<center><img src=\"output/book15.jpg\" />图12 Mask R-CNN 模型总览</center>\n",
    "\n",
    "  \n",
    "  图13是论文给出的架构图。\n",
    " \n",
    "　　<center><img src=\"output/book12.jpg\" />图13 R-CNN 系统架构</center>\n",
    " \n",
    "   它的创新之处在于：\n",
    "\n",
    "　　第一，**三剑合一**：增加目标分割分支，将目标分类、检测、分割算法统一到一个框架下；\n",
    "  \n",
    "　　第二，**ROI精准对齐**：提出了一个简单的，量化无关的层，称为RoIAlign，可保留精确的空间位置；\n",
    "  \n",
    "　　第三，**掩码解耦**：Mask R-CNN为每个类独立地预测二进制掩码，消除了跨类竞争。 \n",
    "  \n",
    "　　所以，Mask R-CNN是在Faster R-CNN的基础上演化而来的，粗略地说，它就是Faster R-CNN和FCN的结合，前者负责目标检测（分类标签,确定位置），后者负责目标分割 (确定掩码）。其关系如图14：\n",
    "  \n",
    "<center><img src=\"output/book13.jpg\" />图14 Mask R-CNN ＝ Faster R-CNN + FCN</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.4.3 实现\n",
    "\n",
    "　　Mask R-CNN在github上有多种开源版本，官方用Caffe2实现，地址位于https://github.com/facebookresearch/Detectron　，我们选取 https://github.com/delldu/MaskRCNN 来分析。选择的理由非常简单：该版本是用**PyTorch**实现的，通过了我们的**测试**。它的安装和使用方法，请参见项目自带的README.md。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4.4.3.1 数据探索\n",
    "\n",
    "　　训练和验证数据集是MS COCO 2014。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 文件概览\n",
    "　　首先，数据集包含两类文件：１.标注文件，位于annotations目录下，json格式；２．图像文件，jpg格式，例子如下："
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "    data\n",
    "    ├── annotations\n",
    "    │   ├── captions_train2014.json\n",
    "    │   ├── captions_val2014.json\n",
    "    │   ├── instances_minival2014.json\n",
    "    │   ├── instances_train2014.json\n",
    "    │   ├── instances_val2014.json\n",
    "    │   ├── instances_valminusminival2014.json\n",
    "    │   ├── person_keypoints_train2014.json\n",
    "    │   └── person_keypoints_val2014.json\n",
    "    ├── test2014\n",
    "    │   ├── COCO_test2014_000000000001.jpg\n",
    "    │   ├── COCO_test2014_000000000014.jpg\n",
    "    │   ├── COCO_test2014_000000000016.jpg\n",
    "    │   ├── COCO_test2014_000000581919.jpg\n",
    "    │   └── COCO_test2014_000000581923.jpg\n",
    "    ├── train2014\n",
    "    │   ├── COCO_train2014_000000000009.jpg\n",
    "    │   ├── COCO_train2014_000000000049.jpg\n",
    "    │   ├── COCO_train2014_000000000061.jpg\n",
    "    │   ├── COCO_train2014_000000581909.jpg\n",
    "    │   └── COCO_train2014_000000581921.jpg\n",
    "    └── val2014\n",
    "    ​    ├── COCO_val2014_000000000042.jpg\n",
    "    ​    ├── COCO_val2014_000000000139.jpg\n",
    "    ​    ├── COCO_val2014_000000000143.jpg\n",
    "    ​    ├── COCO_val2014_000000581913.jpg\n",
    "    ​    └── COCO_val2014_000000581929.jpg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　其次，请重点关注标注文件，因为标注文件是json的紧凑格式，细项堆积在一起，难以阅读，可以使用json工具给予转换，命令如下：\n",
    "\n",
    "`cat data/annotations/instances_valminusminival2014.json | python -m json.tool`\n",
    "\n",
    "　　一个标注文件对应多个图像，它相当一个数据库，通过它可以检索到图像文件的ＩＤ，文件名称，图片包含的目标类别，位置框，掩码等信息。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 访问实战\n",
    "\n",
    "　　和通常PyTorch模型使用数据的要求一样，先构造标准的Dataset类CocoMaskRCNNDataset，它从torchvision.datasets.CocoDetection继承而来，后者从torch.utils.data.Dataset继承而来，需要实现关键函数`__len__()，__getitem__(index)`，分别用于获取样本的总数和单样本的内容，实际内容包括图片数据，标注类别，目标框和掩码，它的复杂性在于：一、分析标注文件，获取基本的标注信息，这部分已由cocoapi实现，请参阅https://github.com/cocodataset/cocoapi.git\n",
    "；二、同步转换图片大小，目标框大小和掩码，并规范化成张量，以适应后续深度网络处理的需要，我们实现了这一部分，核心内容存储在文件data.py中，大家可以尽情探索。\n",
    "\n",
    "　　**重要提示**\n",
    "  \n",
    "　　PyTorch模块nn.Module中的输入一般是BxCxHxW形式的张量(B代表批量，C代表通道，H代表高度，W代表宽度)，所以我们要将CocoMaskRCNNDataset中的图片处理成CxHxW维张量，经过torch.utils.data.DataLoader调用后才能形成BxCxHxW维张量，进入网络运算。实际上初始输入图像的维度是3x1024x1024。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loading annotations into memory...\n",
      "Done (t=5.00s)\n",
      "creating index...\n",
      "index created!\n",
      "Dataset pictures: 35504\n",
      "Dataset memory 712628120 bytes, about 679 M.\n"
     ]
    }
   ],
   "source": [
    "import data\n",
    "import config\n",
    "\n",
    "valid_dataset = data.CocoMaskRCNNDataset(\"data\", \"valminusminival\", 2014, config.Config())\n",
    "valid_dataset.summary()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 改进方向\n",
    "\n",
    "　　可以看出，instances_valminusminival2014.json标注了35504张图片，**值得注意**的是：数据占用内存达到**679Ｍ**！这里可以改进的是：将标注文件和图片转换一次完成，并存入数据库中，学习网络直接从数据库中获取数据，既能减小内存占用，又能加快训练速度。\n",
    "\n",
    "##### 原始标注"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image file ID:  449856 , name:  data/val2014/COCO_val2014_000000449856.jpg\n",
      "Image size: Height x Width =  631 x 640\n"
     ]
    }
   ],
   "source": [
    "valid_dataset.show(4000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/bear.jpg\" />图15 原始标注样本</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 规范输入\n",
    "\n",
    "　　网络输入的图片需要统一变换到3x1024x1024维张量，然后减去ImageNet图像集的颜色均值(RGB[123.7, 116.8, 103.9])。维度变换就是图像缩放：保持图像实质内容长宽比率不变，短边缩放到800像素，长边不超过1024像素，边框不足部分用黑色填充，如下图所示。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image file ID:  449856 , name:  data/val2014/COCO_val2014_000000449856.jpg\n",
      "Image size: Height x Width =  1024 x 1024\n"
     ]
    }
   ],
   "source": [
    "valid_dataset.net_show(4000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/net_bear.jpg\" />图16 缩放图片样本</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 类别空洞\n",
    "　　需要注意的是，COCO数据集的原始类别并不是连续的，例如标识为\"traffic light\"的类别号为84，它超出了数据集类别(含背景共81类)的上限，原因是标注文件类别有些类别(如类别12)不存在，类别空间存在“空洞”，而深度学习网络中的标签一般要求连续，所以需要转换。为此，我们开发了CocoLabel类辅助转换，同时可以获取类别中文名称，具体请参阅data.py中的CocoLabel类。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "traffic light\n",
      "红绿灯\n",
      "74\n",
      "84\n"
     ]
    }
   ],
   "source": [
    "import data\n",
    "\n",
    "print(data.CocoLabel.name(10))\n",
    "print(data.CocoLabel.zh_name(10))\n",
    "print(data.CocoLabel.from_class(84))\n",
    "print(data.CocoLabel.to_class(74))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4.4.3.2 模型分解\n",
    "　　Mask R-CNN包含５个子网：卷积骨干网，特征金字塔网，区域建议网，分类检测网和目标分割网，其中卷积骨干网和特征金字塔网主要用于提取网络特征，区域建议网给出可能包含目标的图像区域，分类检测网完成目标分类和位置检测，目标分割网完成目标分割掩码的计算，５个子网协同工作，共同构成一个深度学习监督网络。下面根据模型总揽图来逐一分解，先介绍相关背景知识，然后说明具体实现。**重要的事要说三遍**，**重要的图也要贴两遍**，所以再贴一遍模型总览图！\n",
    "<center><img src=\"output/book15.jpg\" />图17 Mask R-CNN 模型总览</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 背景知识\n",
    "\n",
    "###### 监督学习模型\n",
    "\n",
    "　　监督学习可以概括成如下模型：\n",
    "\n",
    "　　目标: 学出模型 $Y=f(X)$\n",
    "  \n",
    "　　训练: 通过样本 (X, Y) 训练出$f$。\n",
    "  \n",
    "　　推理: 给定X, 通过$f$计算出Y。\n",
    "\n",
    "　　约束: \n",
    "  \n",
    "　　　　1)维度约束: X, Y 的特征维度固定；\n",
    "    \n",
    "　　　　2)目标约束: 训练时，f(X)与Y尽可能相近，推理时，f(X)尽可能与实际相近。\n",
    "    \n",
    "　　尽管Mask R-CNN网络结构复杂，但最核心的就是这个模型，所以我们后面会按照这个思路来详细剖析。\n",
    "\n",
    "###### 检测框\n",
    "\n",
    "　　检测框，简称框(Box)，是一个矩形，在Mask R-CNN实现中用[y1, x1, y2, x2]表示，左上角坐标为[y1, x1], 右下角为[y2, x2], y从上到下，x从左到右。Mask R-CNN中有各种各样的框，一个锚点是一个框，一个感兴趣区域也是一个框。\n",
    "  \n",
    "###### 绑定框\n",
    "\n",
    "　　绑定框(Bounding Box)对应的是框(Box)的偏移，**和谁绑定？**在建议网中和锚点绑定，在检测网中和ROI(感兴趣区域，由建议网生成)绑定。它的缺省值是[0, 0, 0, 0]，实现中变量名常包含关键字bbox或delta。建议网和检测网输出的检测框采用绑定框，既简化了网络设计，又降低了训练难度，增加了稳定性。设想一下：即使没有任何训练，偏移基本正确(在０附近)！Faster R-CNN把偏移的高度和宽度取对数，还原正常值时取指数，好处是把较大的偏移转换成近似线性变化，对较小偏移几乎没有影响，设计很巧妙！\n",
    "\n",
    "###### ROI\n",
    "\n",
    "　　ROI是感兴趣区域的英文缩写，也一个框，在Mask R-CNN中表示为相对值，4个分量是浮点数，取值位于[0,1]中，通过它截取局部特征，无论大的特征图在空间上如何变化(变大或变小) ，框内的内容实质上都指向同一内容，这是它的美妙之处。\n",
    "  \n",
    "###### 覆盖矩阵\n",
    "　　两个框box1和box2之间可能有重叠，重叠程度常用IoU来度量，即用box1与box2交集的面积除以二者并集的面积。如果有两个框集合，比如A和G，A中的框A<sub>i</sub>与G中的框G<sub>j</sub>的IoU记为s<sub>ij</sub>,则全部s<sub>ij</sub>构成一个矩阵，我们称为覆盖矩阵。例如:\n",
    "\n",
    "|  | G<sub>1</sub>   | G<sub>2</sub> | ... |\n",
    "| ------ | ------ | ------ |\n",
    "| A<sub>1</sub>   | 0.1 | 0.2 |\n",
    "| A<sub>2</sub>   | 0.9 | 0.3 |\n",
    "| A<sub>3</sub>   | 0.2 | 0.3 |\n",
    "|... | | |\n",
    "\n",
    "###### 全卷积网\n",
    "　　全卷积网的英文缩写是FCN，全称是Fully Convolutional Network。\n",
    "  \n",
    "　　通常CNN网络在卷积层之后会接上若干个全连接层，将卷积层产生的特征图映射成一个固定长度的特征向量。以AlexNet为代表的经典结构适合于图像的分类和回归任务，因为它们最后都期望得到整个输入图像的一个数值描述（概率），比如AlexNet的ImageNet模型输出1000维的向量表示输入图像属于每一类的概率。图18中的猫，输入AlexNet，得到一个长为1000的输出向量， 表示输入图像属于每一类的概率，其中在“tabby cat”这一类概率最高。\n",
    "  <center><img src=\"output/book23.jpg\" />图18 CNN分类示意图</center>\n",
    "  \n",
    "　　FCN是把所有的全连接层换成卷积层，原来只能输出一个类别的网络就可以在特征图的每一个像素处输出一个分类结果，这样把原来分类的向量，变成了分类的特征图（热图），原理示意如下：\n",
    "  <center><img src=\"output/book24.jpg\" />图19 CNN到FCN的改进</center>\n",
    "  \n",
    "　　因此，FCN可以接受任意尺寸的输入图像，采用反卷积层对最后卷积层的特征上采样，使它恢复到与输入图像相同的尺寸，可以对每个像素产生预测值，因为它保留了原始输入图像中的空间信息，所以最后可以在上采样的特征图上进行逐像素分类，从而解决了图像语义分割的问题(只能区分类别，无法区分实例)。图20是FCN的作者给出的用于语义分割的全卷积网络的结构图：\n",
    "  <center><img src=\"output/book25.jpg\" />图20 FCN分割结构图</center>\n",
    "\n",
    "　　为了空间映射，需要放大特征图，把特征图恢复到原图的大小，常用的方法有三种：一是反卷积（Deconvolution），这里反卷积不是真正意义的“卷积逆”，用“转置卷积”（Transposed Convolution）代替更合适，但原论文中用的是Deconvolution，本章还是沿用这个名词，它的主要目的是实现上采样；二是膨胀卷积（Dilated Convolution），或者更形象地称为带洞卷积，它可以使特征图视野变大，但不增加计算量，对图像分割，有利于提取全局信息，使得分割更准确（如图21）；三是双线性插值上采样，它是一个古老的技术，计算公式如图22，在Mask R-CNN的ROI对齐上使用了这一技术，能显著地提高分割的精度，是Mask R-CNN的核心模块。\n",
    "  <center><img src=\"output/book26.jpg\" />图21 膨胀卷积示意图</center>\n",
    "  <center><img src=\"output/bilinear.png\" />图22 双线性插值</center>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 卷积骨干网\n",
    "\n",
    "　　项目中Mask R-CNN的卷积骨干网使用的是残差网络Resnet101，网络在前向传播中，特征图会逐步变小。根据特征图的大小，可分为５个阶段，每个阶段特征图大小不变。如果把第i个阶段的最后残差激活输出记为$C_i$，最终有５个输出${C_1，C_2，C_3，C_4，C_5}$，相对于输入图像3x1024x1024，像素步长分别为{2，4，8，16，32}。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "############################################################\n",
    "#  Resnet Graph\n",
    "############################################################\n",
    "class Bottleneck(nn.Module):\n",
    "    expansion = 4\n",
    "\n",
    "    def __init__(self, inplanes, planes, stride=1, downsample=None):\n",
    "        super(Bottleneck, self).__init__()\n",
    "        self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, stride=stride)\n",
    "        self.bn1 = nn.BatchNorm2d(planes, eps=0.001, momentum=0.01)\n",
    "        self.padding2 = SamePad2d(kernel_size=3, stride=1)\n",
    "        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3)\n",
    "        self.bn2 = nn.BatchNorm2d(planes, eps=0.001, momentum=0.01)\n",
    "        self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1)\n",
    "        self.bn3 = nn.BatchNorm2d(planes * 4, eps=0.001, momentum=0.01)\n",
    "        self.relu = nn.ReLU(inplace=True)\n",
    "        self.downsample = downsample\n",
    "        self.stride = stride\n",
    "\n",
    "    def forward(self, x):\n",
    "        residual = x\n",
    "\n",
    "        out = self.conv1(x)\n",
    "        out = self.bn1(out)\n",
    "        out = self.relu(out)\n",
    "\n",
    "        out = self.padding2(out)\n",
    "        out = self.conv2(out)\n",
    "        out = self.bn2(out)\n",
    "        out = self.relu(out)\n",
    "\n",
    "        out = self.conv3(out)\n",
    "        out = self.bn3(out)\n",
    "\n",
    "        if self.downsample is not None:\n",
    "            residual = self.downsample(x)\n",
    "\n",
    "        out += residual\n",
    "        out = self.relu(out)\n",
    "\n",
    "        return out\n",
    "\n",
    "\n",
    "class ResNet(nn.Module):\n",
    "    def __init__(self, architecture, stage5=False):\n",
    "        super(ResNet, self).__init__()\n",
    "        assert architecture in [\"resnet50\", \"resnet101\"]\n",
    "        self.inplanes = 64\n",
    "        self.layers = [3, 4, {\"resnet50\": 6, \"resnet101\": 23}[architecture], 3]\n",
    "        self.block = Bottleneck\n",
    "        self.stage5 = stage5\n",
    "\n",
    "        self.C1 = nn.Sequential(\n",
    "            nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3),\n",
    "            nn.BatchNorm2d(64, eps=0.001, momentum=0.01),\n",
    "            nn.ReLU(inplace=True),\n",
    "            SamePad2d(kernel_size=3, stride=2),\n",
    "            nn.MaxPool2d(kernel_size=3, stride=2),\n",
    "        )\n",
    "        self.C2 = self.make_layer(self.block, 64, self.layers[0])\n",
    "        self.C3 = self.make_layer(self.block, 128, self.layers[1], stride=2)\n",
    "        self.C4 = self.make_layer(self.block, 256, self.layers[2], stride=2)\n",
    "        if self.stage5:\n",
    "            self.C5 = self.make_layer(\n",
    "                self.block, 512, self.layers[3], stride=2)\n",
    "        else:\n",
    "            self.C5 = None\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.C1(x)\n",
    "        x = self.C2(x)\n",
    "        x = self.C3(x)\n",
    "        x = self.C4(x)\n",
    "        if self.C5:\n",
    "            x = self.C5(x)\n",
    "        return x\n",
    "\n",
    "    def stages(self):\n",
    "        return [self.C1, self.C2, self.C3, self.C4, self.C5]\n",
    "\n",
    "    def make_layer(self, block, planes, blocks, stride=1):\n",
    "        downsample = None\n",
    "        if stride != 1 or self.inplanes != planes * block.expansion:\n",
    "            downsample = nn.Sequential(\n",
    "                nn.Conv2d(\n",
    "                    self.inplanes,\n",
    "                    planes * block.expansion,\n",
    "                    kernel_size=1,\n",
    "                    stride=stride),\n",
    "                nn.BatchNorm2d(\n",
    "                    planes * block.expansion, eps=0.001, momentum=0.01),\n",
    "            )\n",
    "\n",
    "        layers = []\n",
    "        layers.append(block(self.inplanes, planes, stride, downsample))\n",
    "        self.inplanes = planes * block.expansion\n",
    "        for i in range(1, blocks):\n",
    "            layers.append(block(self.inplanes, planes))\n",
    "\n",
    "        return nn.Sequential(*layers)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 特征扩边\n",
    "  \n",
    "　　您可能也和我当初一样好奇，如何才能做到每个阶段的特征图大小一样？答案很简单，就是在卷积、池化前，先对特征扩边，如下图所示：\n",
    "  \n",
    "  <center><img src=\"output/samepad.jpg\"　/>图23 特征扩边</center>\n",
    "  \n",
    "　　实做中“特征扩边”是用如下代码定义的(model.py)：\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class SamePad2d(nn.Module):\n",
    "    \"\"\"'SAME' padding.\"\"\"\n",
    "\n",
    "    def __init__(self, kernel_size, stride):\n",
    "        \"\"\"Init.\"\"\"\n",
    "        super(SamePad2d, self).__init__()\n",
    "        self.kernel_size = kernel_size\n",
    "        self.stride = stride\n",
    "\n",
    "    def forward(self, input):\n",
    "        \"\"\"Keep feature size not changing after CNN.\"\"\"\n",
    "        in_width = input.size(2)\n",
    "        in_height = input.size(3)\n",
    "        out_width = math.ceil(float(in_width) / float(self.stride))\n",
    "        out_height = math.ceil(float(in_height) / float(self.stride))\n",
    "        pad_along_width = (\n",
    "            (out_width - 1) * self.stride + self.kernel_size - in_width)\n",
    "        pad_along_height = (\n",
    "            (out_height - 1) * self.stride + self.kernel_size - in_height)\n",
    "        pad_left = math.floor(pad_along_width / 2)\n",
    "        pad_top = math.floor(pad_along_height / 2)\n",
    "        pad_right = pad_along_width - pad_left\n",
    "        pad_bottom = pad_along_height - pad_top\n",
    "        return F.pad(input, (pad_left, pad_right, pad_top, pad_bottom), 'constant', 0)\n",
    "\n",
    "    def __repr__(self):\n",
    "        \"\"\"Dump class name.\"\"\"\n",
    "        return self.__class__.__name__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 特征金字塔网\n",
    "\n",
    "###### 背景介绍\n",
    "　　特征金字塔网络的英文全称是Feature Pyramid Network，缩写为FPN。\n",
    "  \n",
    "　　多尺度目标检测是机器视觉领域的开放课题，极具挑战。直观上看，在图像金字塔上提取特征来构建特征金字塔，可以检测大范围内尺度变化的目标，具有一定的尺度不变性，但是它有很大的局限性: 首先运算时间成倍增加，其次实践证明在训练深度网络时太吃显存，难以使用。因此FPN另辟蹊径，通过融合卷积网输出的多层特征图来构建金字塔 ，在金字塔的每层特征图上来检测目标，推断类别。\n",
    "\n",
    "<center><img src=\"output/book16.jpg\" />图24 FPN结构图</center>\n",
    "\n",
    "　　我们把上图左边的三个特征图称为$C_1、C_2、C_3$，它来自卷积骨干网，右边的三个特征图称为$P_1、P_2、P_3$，是FPN的输出。实际上，这是FPN结构的一个简化图，实做稍复杂，有５层，如图25。它有自下而上、自上而下、横向连接三条路径：\n",
    "\n",
    "<center><img src=\"output/book17.jpg\" />图25 实做FPN结构图</center>\n",
    "\n",
    "　　1.自下而上：就是卷积网的前馈输出，该算法计算出不同比例的特征图，缩放比例为2，具体是Resnet101中的{$C_2，C_3，C_4，C_5$}，因为内存占用较大，实做中没有把$C_1$映射到$P_1$。\n",
    "  \n",
    "　　2.自上而下：首先将空间低分辨率的特征图2倍上采样，其次与空间高分辨率的特征图（横向连接后的结果）逐元素相加合并，最后，在合并图上附加一个3×3卷积来生成最终的特征映射，目的是减少上采样的混叠效应，最终的特征映射集称为{$P_2，P_3，P_4，P_5$}，分别对应于{$C_2，C_3，C_4，C_5$}，$C_i$与$P_i$之间具有相同的维度；为了目标前景检测，$P_5$还下采样到$P_6$。\n",
    "  \n",
    "　　3.横向连接：主要实施1x1或3x3卷积，目的是将$C_i$的输出转换成与$P_i$相同的维度。\n",
    "  \n",
    "　　由于金字塔的所有层共享分类器和回归器，因此所有特征图中选取相同的输出通道(实做选取256）。两组特征{$P_2，P_3，P_4，P_5，P_6$}，{$P_2，P_3，P_4，P_5$}分别作为区域建议网、目标检测网和目标分割网的输入特征来源，完成目标检测和分割任务。\n",
    "\n",
    "###### 特征维度\n",
    "\n",
    "|  | C  | H | W |\n",
    "| ------ | ------ | ------ | ------ |\n",
    "| P<sub>2</sub>   | 256 | 256 |256|\n",
    "| P<sub>3</sub>   | 256 | 128 |128|\n",
    "| P<sub>4</sub>   | 256 | 64 |64|\n",
    "| P<sub>5</sub>   | 256 | 32 |32|\n",
    "| P<sub>6</sub>   | 256 | 16 |16|\n",
    "\n",
    "　　这里的描述看起来抽象复杂，但FPN的定义本身可以用如下代码(model.py)简洁地完成。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class FPN(nn.Module):\n",
    "\n",
    "\"\"\"See the paper \"Feature Pyramid Networks for Object Detection\" for more details.\"\"\"\n",
    "\n",
    "def __init__(self, C1, C2, C3, C4, C5, out_channels):\n",
    "    \"\"\"FPN Init.\"\"\"\n",
    "    super(FPN, self).__init__()\n",
    "    self.out_channels = out_channels\n",
    "    self.C1 = C1\n",
    "    self.C2 = C2\n",
    "    self.C3 = C3\n",
    "    self.C4 = C4\n",
    "    self.C5 = C5\n",
    "    self.P6 = nn.MaxPool2d(kernel_size=1, stride=2)\n",
    "    self.P5_conv1 = nn.Conv2d(2048, self.out_channels, kernel_size=1, stride=1)\n",
    "    self.P5_conv2 = nn.Sequential(\n",
    "        SamePad2d(kernel_size=3, stride=1),\n",
    "        nn.Conv2d(self.out_channels, self.out_channels, kernel_size=3, stride=1),\n",
    "    )\n",
    "    self.P4_conv1 = nn.Conv2d(\n",
    "        1024, self.out_channels, kernel_size=1, stride=1)\n",
    "    self.P4_conv2 = nn.Sequential(\n",
    "        SamePad2d(kernel_size=3, stride=1),\n",
    "        nn.Conv2d(self.out_channels, self.out_channels, kernel_size=3, stride=1),\n",
    "    )\n",
    "    self.P3_conv1 = nn.Conv2d(\n",
    "        512, self.out_channels, kernel_size=1, stride=1)\n",
    "    self.P3_conv2 = nn.Sequential(\n",
    "        SamePad2d(kernel_size=3, stride=1),\n",
    "        nn.Conv2d(self.out_channels, self.out_channels, kernel_size=3, stride=1),\n",
    "    )\n",
    "    self.P2_conv1 = nn.Conv2d(256, self.out_channels, kernel_size=1, stride=1)\n",
    "    self.P2_conv2 = nn.Sequential(\n",
    "        SamePad2d(kernel_size=3, stride=1),\n",
    "        nn.Conv2d(self.out_channels, self.out_channels, kernel_size=3, stride=1),\n",
    "    )\n",
    "\n",
    "def forward(self, x):\n",
    "    x = self.C1(x)\n",
    "    x = self.C2(x)\n",
    "    c2_out = x\n",
    "    x = self.C3(x)\n",
    "    c3_out = x\n",
    "\n",
    "    x = self.C4(x)\n",
    "    c4_out = x\n",
    "    x = self.C5(x)\n",
    "    p5_out = self.P5_conv1(x)\n",
    "\n",
    "    p4_out = self.P4_conv1(c4_out) + F.interpolate(p5_out, scale_factor=2)\n",
    "    p3_out = self.P3_conv1(c3_out) + F.interpolate(p4_out, scale_factor=2)\n",
    "    p2_out = self.P2_conv1(c2_out) + F.interpolate(p3_out, scale_factor=2)\n",
    "\n",
    "    p5_out = self.P5_conv2(p5_out)\n",
    "    p4_out = self.P4_conv2(p4_out)\n",
    "    p3_out = self.P3_conv2(p3_out)\n",
    "    p2_out = self.P2_conv2(p2_out)\n",
    "\n",
    "    p6_out = self.P6(p5_out)\n",
    "\n",
    "    return [p2_out, p3_out, p4_out, p5_out, p6_out]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络输出"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FPN(\n",
      "  (C1): Sequential(\n",
      "    (0): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3))\n",
      "    (1): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace)\n",
      "    (3): SamePad2d\n",
      "    (4): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  )\n",
      "  (C2): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "  )\n",
      "  (C3): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(2, 2))\n",
      "      (bn1): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2))\n",
      "        (1): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (3): Bottleneck(\n",
      "      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "  )\n",
      "  (C4): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(2, 2))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2))\n",
      "        (1): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (3): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (4): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (5): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (6): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (7): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (8): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (9): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (10): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (11): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (12): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (13): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (14): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (15): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (16): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (17): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (18): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (19): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (20): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (21): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (22): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(1024, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "  )\n",
      "  (C5): Sequential(\n",
      "    (0): Bottleneck(\n",
      "      (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(2, 2))\n",
      "      (bn1): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(2048, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "      (downsample): Sequential(\n",
      "        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2))\n",
      "        (1): BatchNorm2d(2048, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      )\n",
      "    )\n",
      "    (1): Bottleneck(\n",
      "      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(2048, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "    (2): Bottleneck(\n",
      "      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn1): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (padding2): SamePad2d\n",
      "      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1))\n",
      "      (bn2): BatchNorm2d(512, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))\n",
      "      (bn3): BatchNorm2d(2048, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)\n",
      "      (relu): ReLU(inplace)\n",
      "    )\n",
      "  )\n",
      "  (P6): MaxPool2d(kernel_size=1, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  (P5_conv1): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "  (P5_conv2): Sequential(\n",
      "    (0): SamePad2d\n",
      "    (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "  )\n",
      "  (P4_conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "  (P4_conv2): Sequential(\n",
      "    (0): SamePad2d\n",
      "    (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "  )\n",
      "  (P3_conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "  (P3_conv2): Sequential(\n",
      "    (0): SamePad2d\n",
      "    (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "  )\n",
      "  (P2_conv1): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "  (P2_conv2): Sequential(\n",
      "    (0): SamePad2d\n",
      "    (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1))\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "from PIL import Image\n",
    "import config as configlib\n",
    "import model as modellib\n",
    "import data as datalib\n",
    "import tools.canvas as convaslib\n",
    "\n",
    "\n",
    "def load_model():\n",
    "    \"\"\"Loading Mask R-CNN for coco.\"\"\"\n",
    "    config = configlib.CocoConfig()\n",
    "    net = modellib.MaskRCNN(config=config, model_dir=\"logs\")\n",
    "    net.load_weights(\"models/mask_rcnn_coco.pth\")\n",
    "    net.eval()\n",
    "\n",
    "    net.training = False\n",
    "    net = net.cuda()\n",
    "\n",
    "    return net\n",
    "\n",
    "config = configlib.Config()\n",
    "model = load_model()\n",
    "image_file = \"images/car58a54312d.jpg\"\n",
    "\n",
    "\n",
    "def show_fpn():\n",
    "    \"\"\"Ouput fpn network.\"\"\"\n",
    "    print(model.fpn)\n",
    "\n",
    "show_fpn()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 可视化P2\n",
    "　　FPN接受规范化的图片张量，调用Resnet101计算$C_1,C_2,C_3,C_4,C_5$,输出$P_2,P_3,P_4,P_5,P_6$。$P_2$是1x256x256x256(BxCxHxW格式)维张量，我们选取$P_2$的前16个二维张量(256x256)，将它可视化，输出到图26。可以看出，P2每个通道都保留了输入图的基本信息，从信息论的角度看，冗余较大，似乎可以简化，这应当是简化网络结构的有用线索。  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def show_image():\n",
    "    \"\"\"Show network input image.\"\"\"\n",
    "    image = Image.open(image_file)\n",
    "    image, scale, cropbox = datalib.encode_image(image, 800, 1024)\n",
    "    image.show()\n",
    "\n",
    "show_image()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/net_car.jpg\" />图26 测试网络图</center>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def show_p2():\n",
    "    \"\"\"Show P2.\"\"\"\n",
    "    image = Image.open(image_file)\n",
    "    image, scale, cropbox = datalib.encode_image(image, 800, 1024)\n",
    "    normalized_image = datalib.normalize_image(image, config.MEAN_PIXEL)\n",
    "    normalized_image.unsqueeze_(0)\n",
    "    normalized_image = normalized_image.cuda()\n",
    "    [p2, p3, p4, p5, p6] = model.fpn(normalized_image)\n",
    "    some_of_p2 = p2[:, :16, :, :]\n",
    "    convaslib.tensor_show(some_of_p2)\n",
    "\n",
    "show_p2()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/net_p2.jpg\" />图27 P2可视图</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 区域建议网\n",
    "\n",
    "###### 背景介绍\n",
    "\n",
    "　　区域建议网的英文全称是Region Proposal Network，简称RPN，来自论文《Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks》，网络结构如图27：\n",
    "  \n",
    "  <center><img src=\"output/book18.jpg\" />图28 RPN结构图</center>\n",
    " \n",
    "　　对于目标检测任务，快速生成可能的检测框是核心任务，经典检测方法都非常耗时，如滑动窗口结合图像金字塔的方法，或选择搜索(Selective Search)方法，Faster R-CNN则直接用RPN生成检测框，极大地提升了检测框的生成速度。\n",
    "\n",
    "　　RPN网络的作用是输入特征图，输出一批候选矩形，每个矩形有两个分数，分别是前景（包含目标）和背景（不含目标）的概率。为了完成这个任务，我们按照1024x1024的尺寸，在空间上设置密密麻麻的候选框，每个框有固定的大小和位置，形象地称为锚点，如果有K个框(锚点)，就有２K个分数，４Ｋ个坐标位置。\n",
    "  \n",
    "  　<center><img src=\"output/book19.jpg\" />图29 锚点分布示意图</center>\n",
    "\n",
    "　　具体实现中，锚点有三种长宽比，分别为(sqrt(2)/2, 1.0, sqrt(2))，对应长形锚，方形锚，扁形锚，三种锚的中心点重合，其中方形锚的大小分别为{32, 64, 128,256, 512}共５种，这里sqrt(x)表示平方根函数，如图30所示，三种大小，三种形状，共９个锚点。之所以选择不同形状的锚点，是为了更好地与目标匹配，这与经典目标检测中选用各种各样的Haar算子有异曲同工之妙。当然选择的形状越多，匹配效果就越好，考虑到算力限制，实做中做了取舍。\n",
    "<center><img src=\"output/book20.jpg\" />图30 锚点形状示意图</center>\n",
    "\n",
    "###### NMS算法\n",
    "   \n",
    "　　由于锚点相互重叠，因此候选框最终可能也会在同一目标上重复。为此，我们使用了非极大抑制(NMS)算法来去重，步骤如下：\n",
    "  <center><img src=\"output/book21.jpg\" />图31 NMS原始框图</center>\n",
    "\n",
    "１．将所有候选框的得分进行排序，选中最高分及其所对应的框（标为红色）；\n",
    "  <center><img src=\"output/book22.jpg\" />图32 NMS选择过程示意图</center>\n",
    "\n",
    "２．遍历其余的框，如果它和当前最高得分框的重叠面积大于一定的阈值，我们将其删除。\n",
    "\n",
    "３．从没有处理的框中继续选择一个得分最高的，重复上述过程，得到最终结果。\n",
    "  <center><img src=\"output/nms.png\" />图33 NMS结果</center>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络模型\n",
    "\n",
    "　　形式上，我们把区域建议网描述成：\n",
    "\n",
    "　　$class\\_logits, class, bbox = RPN([P_2, P_3, P_4, P_5, P_6])$\n",
    "  \n",
    "　　这里:\n",
    "  \n",
    "  　　　class是分类概率，共两个类别：背景和前景；\n",
    "\n",
    "  　　　$class\\_logits$是分类的回归值，在模型训练时用于评估损失，所以class等于$softmax(class\\_logits)$。\n",
    "\n",
    "　　把它的最小核心描述成：　$Y = rpn(X)$\n",
    "  \n",
    "　　**我们的任务就是要回答数据X，目标Ｙ和函数rpn是什么？**\n",
    "   \n",
    "　　答案是：X是锚点特征向量(512维)，Ｙ是锚点标签(2值)和锚点绑定框，rpn就是两个卷积函数(用全卷积代替全连接，作用类似线性回归)，这些满足了监督学习模型的约束条件，它是一个全卷积网。\n",
    "  \n",
    "　　我们以类别判断为例，画出获取X，Y的流程："
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$rpn$ 获取数据X的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "image=>inputoutput: 图片 \n",
    "fpn=>operation: FPN前向计算\n",
    "p2p6=>inputoutput: P2-P6特征\n",
    "conv=>operation: 3x3卷积\n",
    "x=>inputoutput: X：锚点特征向量(512维)\n",
    "e=>end: 结束\n",
    "\n",
    "s->image(right)->fpn(right)->p2p6(right)->conv(right)->x->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$rpn$ 获取目标Y的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "gtboxes=>inputoutput: 标注真值框 \n",
    "overlaps=>operation: 锚点覆盖计算\n",
    "sample=>operation: 平衡抽样(正负样本)\n",
    "y=>inputoutput: Y：锚点标签(2值，共３对)\n",
    "e=>end: 结束\n",
    "\n",
    "s->gtboxes(right)->overlaps(right)->sample(right)->y->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　RPN的核心围绕锚点展开，体现在三个方面：\n",
    "    \n",
    "　　一、特征$P_2,P_3,P_4,P_5,P_6$的空间维度与锚点对应，特征提取以锚点为支撑，即同一空间位置有3个锚点，对应一个512维特征向量和3对标签数据，所以RPN网络定义中对分类最后卷积层的输出通道是6，对绑定框回归输出通道是12，**特别注意：**分类和绑定框回归，共享了同一特征向量。请参考以下锚点的分布统计表和后面的RPN网络定义：\n",
    "\n",
    "|  | 锚类型 | H | W |小计|\n",
    "| ------ | ------ | ------ | ------ | ------ |\n",
    "| P<sub>2</sub>   | 3 | 256 |256|196608|\n",
    "| P<sub>3</sub>   | 3 | 128 |128|49152|\n",
    "| P<sub>4</sub>   | 3 | 64 |64|12288|\n",
    "| P<sub>5</sub>   | 3 | 32 |32|3072|\n",
    "| P<sub>6</sub>   | 3 | 16 |16|768|\n",
    "| 总计 |  |  ||261888|\n",
    "\n",
    "　　二、模型训练时，根据锚点与标注框之间的覆盖率(覆盖矩阵)，抽取正负样本；\n",
    "    \n",
    "　　三、模型推理时，給锚点打分，选取高分锚点及其检测框(ROI格式)推荐输出。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class RPN(nn.Module):\n",
    "    \"\"\"Builds the model of Region Proposal Network.\"\"\"\n",
    "\n",
    "    def __init__(self, anchors_per_location, anchor_stride, depth):\n",
    "        super(RPN, self).__init__()\n",
    "        # anchors_per_location = 3\n",
    "        self.anchor_stride = anchor_stride  # 1\n",
    "        self.depth = depth  # 256\n",
    "\n",
    "        self.padding = SamePad2d(kernel_size=3, stride=self.anchor_stride)\n",
    "        self.conv_shared = nn.Conv2d(self.depth, 512, kernel_size=3, stride=self.anchor_stride)\n",
    "        self.relu = nn.ReLU(inplace=True)\n",
    "        self.conv_class = nn.Conv2d(512, 2 * anchors_per_location, kernel_size=1, stride=1)\n",
    "        self.softmax = nn.Softmax(dim=2)\n",
    "        self.conv_bbox = nn.Conv2d(512, 4 * anchors_per_location, kernel_size=1, stride=1)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.relu(self.conv_shared(self.padding(x)))\n",
    "\n",
    "        # Anchor Score. [batch, anchors per location * 2, height, width].\n",
    "        rpn_class_logits = self.conv_class(x)\n",
    "\n",
    "        # Reshape to [batch, 2, anchors]\n",
    "        rpn_class_logits = rpn_class_logits.permute(0, 2, 3, 1)\n",
    "        rpn_class_logits = rpn_class_logits.contiguous()\n",
    "        rpn_class_logits = rpn_class_logits.view(x.size()[0], -1, 2)\n",
    "\n",
    "        # Softmax on last dimension of BG/FG.\n",
    "        rpn_class = self.softmax(rpn_class_logits)\n",
    "\n",
    "        # Bounding box refinement. [batch, H, W, anchors per location, depth]\n",
    "        # where depth is [x, y, log(w), log(h)]\n",
    "        rpn_bbox = self.conv_bbox(x)\n",
    "\n",
    "        # Reshape to [batch, 4, anchors]\n",
    "        rpn_bbox = rpn_bbox.permute(0, 2, 3, 1)\n",
    "        rpn_bbox = rpn_bbox.contiguous()\n",
    "        rpn_bbox = rpn_bbox.view(x.size()[0], -1, 4)\n",
    "\n",
    "        return [rpn_class_logits, rpn_class, rpn_bbox]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络输出"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RPN(\n",
      "  (padding): SamePad2d\n",
      "  (conv_shared): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1))\n",
      "  (relu): ReLU(inplace)\n",
      "  (conv_class): Conv2d(512, 6, kernel_size=(1, 1), stride=(1, 1))\n",
      "  (softmax): Softmax()\n",
      "  (conv_bbox): Conv2d(512, 12, kernel_size=(1, 1), stride=(1, 1))\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "print(model.rpn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 再论锚点\n",
    "\n",
    "　　锚点到底长什么样？\n",
    "\n",
    "　　我们知道，在$P_2,P_3,P_4,P_5,P_6$特征上都要建立锚点，如果网络输入图片大小为1024x1024, 总计有261888个锚点。下面以$P_3$为例，将锚点的中心点在网络图上标出。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def draw_anchors():\n",
    "    \"\"\"Draw anchors base on P3 dimension.\"\"\"\n",
    "    import data\n",
    "    import utils\n",
    "    image = Image.open(\"images/car58a54312d.jpg\")\n",
    "    image, scale, crop = data.encode_image(image, 800, 1024)\n",
    "    anchors = utils.create_anchors(scales=64, ratios=[0.5, 1, 2], shape=[128, 128], feature_stride=8, anchor_stride=1)\n",
    "    data.draw_anchors(image, anchors)\n",
    "\n",
    "draw_anchors()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/anchors.jpg\" />图34 P3锚点分布图</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### RPN NMS\n",
    "\n",
    "　　如果不做筛选，RPN共有261888个锚点，所以有261888个建议框，根据RPN检测前景的得分结果，采用NMS抑制后，建议框大大减少，下面是测试代码和检测结果。这些建议框规范化后变成ROI，作为参数，会送入检测网络和分割网络。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def refine_rpn():\n",
    "    \"\"\"Refine rpn.\"\"\"\n",
    "    image = Image.open(image_file)\n",
    "    image, scale, cropbox = datalib.encode_image(image, 800, 1024)\n",
    "\n",
    "    normalized_image = datalib.normalize_image(image, config.MEAN_PIXEL)\n",
    "    normalized_image.unsqueeze_(0)\n",
    "    normalized_image = normalized_image.cuda()\n",
    "    [p2, p3, p4, p5, p6] = model.fpn(normalized_image)\n",
    "\n",
    "    rpn_class_logits, rpn_class, rpn_bbox = model.rpn_detect([p2, p3, p4, p5, p6])\n",
    "    rpn_rois = model.rpn_refine(rpn_class, rpn_bbox)\n",
    "    rpn_rois.squeeze_(0)\n",
    "\n",
    "    boxes = datalib.boxes_scale(rpn_rois, [1024, 1024, 1024, 1024])\n",
    "    datalib.blend_image(image, None, boxes, masks=None, scores=None).show()\n",
    "\n",
    "refine_rpn()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/rpn_first.jpg\" />图35 RPN建议框</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 目标检测网"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### ROI 对齐\n",
    "　　检测网和分割网都接受ＦＰＮ输出的特征和RPN输出的ＲＯＩ，对检测网，需要在[P<sub>2</sub>, P<sub>3</sub>, P<sub>4</sub>, P<sub>5</sub>]上提取每个ＲＯＩ对应的特征，并把它统一成256x7x7的特征张量，然后经过２次卷积，拉伸成1024维向量，作为$detect$的Ｘ变量；对分割网，挖补的特征要统一成256x14x14的特征张量，再经过４次卷积和１次反卷积，变成256x28x28的特征张量，最后卷积１次，产生81x28x28维张量，作为$mask$的输入变量X，由此可以看出，检测网和分割网都共享了ＦＰＮ输出的基础特征，差别在于二者细化时，输出结果稍有差异。这里核心算法是利用双线性插值算法进行精确的特征挖补(ROI对齐)，不象Faster R-CNN粗略地只做２次量化，所以它显著地提高了检测和分割的精度，是论文《Mask R-CNN》的重要贡献。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络模型\n",
    "\n",
    "　　形式上，我们把目标检测网描述成：\n",
    "  \n",
    "　　　　$class\\_logits, class, bbox = Detect([P_2, P_3, P_4, P_5], rois)$\n",
    "    \n",
    "　　这里: \n",
    "\n",
    "　　　　rois 是RPN输出的ROI集合；\n",
    "\n",
    "　　　　class是分类概率，共计81(包括背景)；\n",
    "\n",
    "　　　　$class\\_logits$是分类回归值，主要在训练时用于评估损失，即class等于$softmax(class\\_logits)$;\n",
    "\n",
    "　　　　bbox 是检测绑定框。\n",
    "\n",
    "　　同时把它的最小核心描述成：　$Y = detect(X)$\n",
    "  \n",
    "　　**我们的任务就是要回答数据X，目标Ｙ和函数detect是什么？**\n",
    "   \n",
    "　　答案是：X是来自$[P_2,P_3,P_4,P_5]$上的256x7x7特征块，经过一系列卷积变换生成的1024维向量；Ｙ是标注类别(81类)和真值绑定框(通过标注框与ROI计算得到），detect是两个线性回归函数，这些同样满足了监督学习模型的约束条件，但它是一个**全连接网！**\n",
    "  \n",
    "　　以下给出获取X，Y的流程："
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$detect$ 获取数据X的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "p2p5=>inputoutput: P2-P5特征，ROI\n",
    "roialign=>operation: ROI对齐\n",
    "x7x7=>inputoutput: 256x7x7特征图\n",
    "conv=>operation: 卷积变换\n",
    "x=>inputoutput: X：1024维向量\n",
    "e=>end: 结束\n",
    "s->p2p5(right)->roialign(right)->x7x7(right)->conv(right)->x->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$detect$ 获取目标Y的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "y=>inputoutput: Y：标注类别(81类),真值绑定框\n",
    "e=>end: 结束\n",
    "s(right)->y(right)->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Classifier(nn.Module):\n",
    "\n",
    "    def __init__(self, depth, pool_size, image_shape, num_classes):\n",
    "    \n",
    "        super(Classifier, self).__init__()\n",
    "        self.depth = depth\n",
    "        self.pool_size = pool_size\n",
    "        self.image_shape = image_shape\n",
    "        self.num_classes = num_classes\n",
    "        self.conv1 = nn.Conv2d(self.depth, 1024, kernel_size=self.pool_size, stride=1)\n",
    "        self.bn1 = nn.BatchNorm2d(1024, eps=0.001, momentum=0.01)\n",
    "        self.conv2 = nn.Conv2d(1024, 1024, kernel_size=1, stride=1)\n",
    "        self.bn2 = nn.BatchNorm2d(1024, eps=0.001, momentum=0.01)\n",
    "        self.relu = nn.ReLU(inplace=True)\n",
    "\n",
    "        self.linear_class = nn.Linear(1024, num_classes)\n",
    "        self.softmax = nn.Softmax(dim=1)\n",
    "\n",
    "        self.linear_bbox = nn.Linear(1024, num_classes * 4)\n",
    "\n",
    "    def forward(self, x, rois):\n",
    "        x = roi_align([rois] + x, self.pool_size, self.image_shape)\n",
    "        x = self.conv1(x)\n",
    "        x = self.bn1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.conv2(x)\n",
    "        x = self.bn2(x)\n",
    "        x = self.relu(x)\n",
    "\n",
    "        x = x.view(-1, 1024)\n",
    "        mrn_class_logits = self.linear_class(x)\n",
    "        mrn_probs = self.softmax(mrn_class_logits)\n",
    "\n",
    "        mrn_bbox = self.linear_bbox(x)\n",
    "        mrn_bbox = mrn_bbox.view(mrn_bbox.size(0), -1, 4)\n",
    "\n",
    "        return [mrn_class_logits, mrn_probs, mrn_bbox]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 检测实例 \n",
    "   \n",
    "　　检测网的输入是FPN的$P_2,P_3,P_4,P_5$特征图和ROI，注意没有**$P_6$**!输出分类目标的概率（得分）和检测框的位置，它同样是根据目标分类的分值，采用NMS算法抑制后的结果。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def detection():\n",
    "    \"\"\"Detection boxes.\"\"\"\n",
    "    image = Image.open(image_file)\n",
    "    image, scale, cropbox = datalib.encode_image(image, 800, 1024)\n",
    "\n",
    "    normalized_image = datalib.normalize_image(image, config.MEAN_PIXEL)\n",
    "    normalized_image.unsqueeze_(0)\n",
    "    normalized_image = normalized_image.cuda()\n",
    "    [p2, p3, p4, p5, p6] = model.fpn(normalized_image)\n",
    "\n",
    "    rpn_class_logits, rpn_class, rpn_bbox = model.rpn_detect([p2, p3, p4, p5, p6])\n",
    "    rpn_rois = model.rpn_refine(rpn_class, rpn_bbox)\n",
    "    rpn_rois.squeeze_(0)\n",
    "\n",
    "    # Detections\n",
    "    mrn_class_logits, mrn_class, mrn_bbox = model.mrn_detect([p2, p3, p4, p5], rpn_rois)\n",
    "    mrn_class_ids, mrn_scores, mrn_boxes = model.mrn_refine(rpn_rois, mrn_class, mrn_bbox, [0, 0, 1024, 1024])\n",
    "    boxes = mrn_boxes.squeeze(0).detach().cpu()\n",
    "    datalib.blend_image(image, None, boxes, masks=None, scores=None).show()\n",
    "\n",
    "\n",
    "detection()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<center><img src=\"output/rpn_refine.jpg\" />图36 检测实例</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 目标分割网\n",
    "  \n",
    "　　项目实现的卷积骨干网建立在ImageNet Restnet101基础上，输入尺寸最后要求与224x224像素对应，网络结构示意图如下。\n",
    "  <center><img src=\"output/book27.jpg\" />图37 Mask R-CNN分割结构示意图</center>\n",
    "  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络模型\n",
    "\n",
    "　　形式上，我们把目标分割网描述成：\n",
    "  \n",
    "　　　　$masks = Mask([P_2, P_3, P_4, P_5]，rois)​$\n",
    "\n",
    "　　这里:\n",
    "  \n",
    "　　rois 是RPN输出的ROI集合；\n",
    "  \n",
    "　　masks 是81x28x28维的分割掩码集合。\n",
    "\n",
    "　　同时把它的最小核心描述成：　$Y = mask(X)$\n",
    "\n",
    "　　**我们的任务同样是要回答数据X，目标Ｙ和函数mask是什么？**\n",
    "   \n",
    "　　答案是：X是81x28x28维张量，它来自$[P_2,P_3,P_4,P_5]$上的256x14x14特征张量(是经过４次卷积，１次反卷积，１次卷积后生成的），Ｙ是源自真值标注的压缩掩码(28x28)，mask是sigmoid函数，这些同样满足了监督学习模型的约束条件，貌似sigmoid函数不需要学习，但是整体上，分割网的学习建立在卷积和反卷积的基础上，所以它还是一个全卷积网。\n",
    "  \n",
    "　　我们同样给出获取X，Y的流程："
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$mask$ 获取数据X的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "p2p5=>inputoutput: P2-P5特征 + ROI\n",
    "roialign=>operation: ROI对齐(双线性插值)\n",
    "conv=>operation: 卷积平滑\n",
    "deconv=>operation: 反卷积\n",
    "x=>inputoutput: X：81x28x28特征张量\n",
    "e=>end: 结束\n",
    "s->p2p5(right)->roialign(right)->conv(right)->deconv(right)->x->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　**$mask$ 获取目标Y的流程**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```flow\n",
    "flow\n",
    "s=>start: 开始 \n",
    "rawmask=>inputoutput: 标注掩码(0/1)\n",
    "normal=>operation: 掩码压缩\n",
    "out=>inputoutput: Y:真值压缩掩码(28x28)\n",
    "e=>end: 结束\n",
    "s(right)->rawmask(right)->normal(right)->out(right)->e\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 网络定义"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Mask(nn.Module):\n",
    "\n",
    "    def __init__(self, depth, pool_size, image_shape, num_classes):\n",
    "    \n",
    "        super(Mask, self).__init__()\n",
    "        self.depth = depth\n",
    "        self.pool_size = pool_size\n",
    "        self.image_shape = image_shape\n",
    "        self.num_classes = num_classes\n",
    "        self.padding = SamePad2d(kernel_size=3, stride=1)\n",
    "        self.conv1 = nn.Conv2d(self.depth, 256, kernel_size=3, stride=1)\n",
    "        self.bn1 = nn.BatchNorm2d(256, eps=0.001)\n",
    "        self.conv2 = nn.Conv2d(256, 256, kernel_size=3, stride=1)\n",
    "        self.bn2 = nn.BatchNorm2d(256, eps=0.001)\n",
    "        self.conv3 = nn.Conv2d(256, 256, kernel_size=3, stride=1)\n",
    "        self.bn3 = nn.BatchNorm2d(256, eps=0.001)\n",
    "        self.conv4 = nn.Conv2d(256, 256, kernel_size=3, stride=1)\n",
    "        self.bn4 = nn.BatchNorm2d(256, eps=0.001)\n",
    "        self.deconv = nn.ConvTranspose2d(256, 256, kernel_size=2, stride=2)\n",
    "        self.conv5 = nn.Conv2d(256, num_classes, kernel_size=1, stride=1)\n",
    "        self.sigmoid = nn.Sigmoid()\n",
    "        self.relu = nn.ReLU(inplace=True)\n",
    "\n",
    "    def forward(self, x, rois):\n",
    "        x = roi_align([rois] + x, self.pool_size, self.image_shape)\n",
    "\n",
    "        x = self.conv1(self.padding(x))\n",
    "        x = self.bn1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.conv2(self.padding(x))\n",
    "        x = self.bn2(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.conv3(self.padding(x))\n",
    "        x = self.bn3(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.conv4(self.padding(x))\n",
    "        x = self.bn4(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.deconv(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.conv5(x)\n",
    "        x = self.sigmoid(x)\n",
    "\n",
    "        return x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 分割实例\n",
    "\n",
    "　　项目提供一条命令用于完成给定图像的分割，举例如下:\n",
    "\n",
    "　　`python predict.py images/car58a54312d.jpg`\n",
    "\n",
    "　　结果如下：\n",
    "  <center><img src=\"output/car58a54312d.jpg\" />图38 分割实例</center>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### 灰度图分割\n",
    "　　Mask R-CNN对灰度图的分割例子如下，效果也很好。\n",
    "<center><img src=\"output/car_gray.png\" />图39 灰度图分割例子</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "#### 4.4.3.3 损失函数\n",
    "##### 区域建议网\n",
    "  \n",
    "　　分类采用的是交叉熵(cross_entropy)函数，建议框回归采用的是L1平滑函数(Smooth L1)。\n",
    "   \n",
    "##### 目标检测网\n",
    "  \n",
    "　　分类采用的是交叉熵(cross_entropy)函数，检测框回归采用的是L1平滑函数(Smooth L1)。\n",
    "  \n",
    "##### 目标分割网\n",
    "  \n",
    "　　分割采用的二值交叉熵(binary_cross_entropy)。\n",
    "  \n",
    "##### 总体损失\n",
    "\n",
    "　　骨干卷积网和特征金字塔网不计算损失，但它们的损失会传递到上述网络中，所以网络总体损失是以上５类损失直接相加，相对简单，我们不再详细介绍。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4.4.3.4 模型训练\n",
    "##### 训练命令\n",
    "\n",
    "　　请使用命令\n",
    "  \n",
    "`./train.sh`\n",
    "\n",
    "　　训练模型。\n",
    "  \n",
    "　　模型超参数都在config.py中设置，有详细的注释，我们在此省略。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "##### 训练输出\n",
    "\n",
    "    Starting at epoch 1. LR=0.001\n",
    "\n",
    "    Epoch 1/40.\n",
    "\n",
    "        1/1000 |----------| 0.1% Complete - loss: 1.34925 - rpn_class_loss: 0.02360 - rpn_bbox_loss: 0.16599 - mrn_class_loss: 0.71810 - mrn_bbox_loss: 0.17466 - mrn_mask_loss: 0.26690\n",
    "\n",
    "        2/1000 |----------| 0.2% Complete - loss: 1.74254 - rpn_class_loss: 0.08960 - rpn_bbox_loss: 0.15880 - mrn_class_loss: 0.74013 - mrn_bbox_loss: 0.43124 - mrn_mask_loss: 0.32277\n",
    "\n",
    "        3/1000 |----------| 0.3% Complete - loss: 0.40288 - rpn_class_loss: 0.00326 - rpn_bbox_loss: 0.04056 - mrn_class_loss: 0.21698 - mrn_bbox_loss: 0.02083 - mrn_mask_loss: 0.12125\n",
    "\n",
    "        4/1000 |----------| 0.4% Complete - loss: 1.49034 - rpn_class_loss: 0.03006 - rpn_bbox_loss: 0.23697 - mrn_class_loss: 0.59642 - mrn_bbox_loss: 0.24462 - mrn_mask_loss: 0.38227\n",
    "\n",
    "        5/1000 |----------| 0.5% Complete - loss: 1.57298 - rpn_class_loss: 0.09510 - rpn_bbox_loss: 0.18876 - mrn_class_loss: 0.53671 - mrn_bbox_loss: 0.28291 - mrn_mask_loss: 0.46950"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4.4.3.5 模型评估\n",
    "\n",
    "##### 评估命令\n",
    "\n",
    "　　请使用命令\n",
    "  \n",
    "  `./eval.sh`\n",
    "  \n",
    "　　评估模型性能。\n",
    "  \n",
    "  　项目采用的是COCO挑战赛的标准评估程序。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 评估结果\n",
    "\n",
    "    Running COCO evaluation on 500 images.\n",
    "    Running per image evaluation...\n",
    "    Evaluate annotation type *bbox*\n",
    "    DONE (t=1.49s).\n",
    "    Accumulating evaluation results...\n",
    "    DONE (t=0.53s).\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.384\n",
    "     Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.577\n",
    "     Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.412\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.189\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.429\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.518\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.328\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.447\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.454\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.223\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.491\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.592\n",
    "    Prediction time: 200.76635813713074. Average 0.40153271627426146/image\n",
    "    Total time:  265.6884126663208\n",
    "    Evaluate annotation type *segm*\n",
    "    DONE (t=1.63s).\n",
    "    Accumulating evaluation results...\n",
    "    DONE (t=0.50s).\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.169\n",
    "     Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.260\n",
    "     Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.181\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.108\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.209\n",
    "     Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.231\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.172\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.239\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.242\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.130\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.265\n",
    "     Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.314\n",
    "    Prediction time: 214.55515885353088. Average 0.42911031770706176/image\n",
    "    Total time:  290.76130080223083"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4.5 本章小节"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "　　图像分割具有极高的难度。Mask R-CNN架构合理，性能强大，是实例分割领域不可多得的佳品，但它没有达到完美，我们提出一些改进方向，作为本章的小结。\n",
    "\n",
    "#### 一、架构改进\n",
    "　　Mask R-CNN即使在GPU的帮助下，也只能达到5fps，离实时还有差距，简化网络架构是一种可能的方向。\n",
    "\n",
    "#### 二、效果改进\n",
    "\n",
    "　　1.边界附近漏分或错分的比例较大，可能的改进方向是引入导向滤波协助分割；\n",
    " \n",
    "　　2.负片识别错误，如图40, 改进的方向是增强数据，将负片样本加入训练集合;\n",
    "<center><img src=\"output/car_negate.png\" />图40 负片分割</center>\n",
    "\n",
    "　　3.小目标识别错误，如图41，把小花朵识别成了运动球，改进的方向是引进场景识别和图像推理。\n",
    "<center><img src=\"output/img_7233.png\" />图41 小目标分割</center>\n",
    "  \n",
    "#### 三、领域扩展\n",
    "\n",
    "　　1.将Mask R-CNN适配到全景分割领域；\n",
    "  \n",
    "　　2.走进3D分割，通过引入深度信息，既能改进目标分割的速度和精度，又能提高应用价值，因为目标的深度信息对\"无人驾驶\"等应用本身具有极高价值，简单地讲，自动驾驶的目标识别就是一个3D全景分割任务。\n",
    "  \n",
    "　　我们接下来的章节会研究3D图像分割。"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Edit Metadata",
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}