初始化代码仓库“

LICENSE

@@ -1,3 +1,21 @@
-Special exception for linking OpenVPN with OpenSSL:
+MIT License
 
-In addition, as a special exception, OpenVPN Technologies, Inc. gives permission to link the code of this program with the OpenSSL Library (or with modified versions of OpenSSL that use the same license as OpenSSL), and distribute linked combinations including the two. You must obey the GNU General Public License in all respects for all of the code used other than OpenSSL. If you modify this file, you may extend this exception to your version of the file, but you are not obligated to do so. If you do not wish to do so, delete this exception statement from your version.
+Copyright (c) 2018 Bugdragon
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,19 +1,32 @@
-#### 从命令行创建一个新的仓库
+# YOLO_v3_PyTorch
+YOLOv3目标检测器的**输出样例**：
+![Image text](https://raw.githubusercontent.com/Bugdragon/YOLO_v3_PyTorch/master/det/det_%E5%B7%A5%E4%BD%9C%E7%BB%86%E8%83%9E1.png)
 
-```bash
-touch README.md
-git init
-git add README.md
-git commit -m "first commit"
-git remote add origin https://testgitea.trustie.net/daiao/YOLO_v3_PyTorch.git
-git push -u origin master
+### 代码实现
+1. 创建 YOLOv3 网络层级☑
+2. 实现网络的前向传播☑
+3. objectness 置信度阈值和非极大值抑制☑
+4. 设计输入和输出管道☑
+5. 在视频/网络摄像头上运行检测器☑
 
-```
+### 背景知识
++ 卷积神经网络的工作原理，包括残差块、跳过连接和上采样；
++ 目标检测、边界框回归、IoU 和非极大值抑制（NMS）；
++ 基础的 PyTorch 使用，会创建简单的神经网络；
++ 阅读 YOLO 三篇论文，了解 YOLO 的工作原理。
 
-#### 从命令行推送已经创建的仓库
+### 版本条件
+* Ubuntu 18.04LTS(64-bit)
+* Python 3.6.5(pip3)
+* torch 0.4.0(cpu)
+* OpenCV 3.4.2
 
-```bash
-git remote add origin https://testgitea.trustie.net/daiao/YOLO_v3_PyTorch.git
-git push -u origin master
+### 安装指南
+* git clone https://github.com/Bugdragon/YOLO_v3_PyTorch.git
+* cd YOLO_v3_PyTorch
+* wget https://pjreddie.com/media/files/yolov3.weights
+* python detect.py
 
-```
+#### tips
+1. 提前将需要检测的图片放入 imgs 文件夹下
+2. 检测结果图片将被保存在 det 文件夹下

cfg/yolov3.cfg

@@ -0,0 +1,789 @@
+[net]
+# Testing
+batch=1
+subdivisions=1
+# Training
+# batch=64
+# subdivisions=16
+width=416
+height=416
+channels=3
+momentum=0.9
+decay=0.0005
+angle=0
+saturation = 1.5
+exposure = 1.5
+hue=.1
+
+learning_rate=0.001
+burn_in=1000
+max_batches = 500200
+policy=steps
+steps=400000,450000
+scales=.1,.1
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=3
+stride=1
+pad=1
+activation=leaky
+
+# Downsample
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+# Downsample
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+# Downsample
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+# Downsample
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+# Downsample
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=2
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[shortcut]
+from=-3
+activation=linear
+
+######################
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=255
+activation=linear
+
+
+[yolo]
+mask = 6,7,8
+anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
+classes=80
+num=9
+jitter=.3
+ignore_thresh = .7
+truth_thresh = 1
+random=1
+
+
+[route]
+layers = -4
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[upsample]
+stride=2
+
+[route]
+layers = -1, 61
+
+
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=512
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=512
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=512
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=255
+activation=linear
+
+
+[yolo]
+mask = 3,4,5
+anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
+classes=80
+num=9
+jitter=.3
+ignore_thresh = .7
+truth_thresh = 1
+random=1
+
+
+
+[route]
+layers = -4
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[upsample]
+stride=2
+
+[route]
+layers = -1, 36
+
+
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=256
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=256
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=256
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=255
+activation=linear
+
+
+[yolo]
+mask = 0,1,2
+anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
+classes=80
+num=9
+jitter=.3
+ignore_thresh = .7
+truth_thresh = 1
+random=1
+

darknet.py

@@ -0,0 +1,329 @@
+from __future__ import division
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.autograd import Variable
+import numpy as np
+from util import *
+
+# 输入
+def get_test_input():
+
+  img = cv2.imread("dog-cycle-car.png")
+  img = cv2.resize(img, (416,416)) # 调整输入图像维度
+  img_ = img[:,:,::-1].transpose((2,0,1)) # BGR -> RGB | H x W x C -> C x H x W
+  img_ = img_[np.newaxis,:,:,:]/255.0 # 添加通道，置0，作为正则通道
+  img_ = torch.from_numpy(img_).float() # 转换成float
+  img_ = Variable(img_) # 转成变量
+
+  return img_
+
+#Takes a cfg file,returns a list of blocks. 
+def parse_cfg(cfgfile):
+
+  file = open(cfgfile, 'r')
+  lines = file.read().split('\n') # store lines in a list
+  lines = [x for x in lines if len(x)>0] # get rid of empty lines
+  lines = [x for x in lines if x[0] != '#'] # get rid of comments
+  lines = [x.rstrip().lstrip() for x in lines] # get rid of whitespaces
+  
+  block = {}
+  blocks = []
+
+  for line in lines:
+    if line[0] == "[": # a new block
+      if len(block) != 0: # not empty
+        blocks.append(block) # add blocks list
+        block = {} # init blocks
+      block["type"] = line[1:-1].rstrip()
+    else:
+      key, value = line.split("=")
+      block[key.rstrip()] = value.lstrip()
+  blocks.append(block)
+
+  return blocks
+
+# 空层
+class EmptyLayer(nn.Module):
+  def __init__(self):
+    super(EmptyLayer, self).__init__() # 调用父类方法
+
+# 定义一个新的DetectionLayer保存检测边界框的锚点
+class DetectionLayer(nn.Module):
+  def __init__(self, anchors):
+    super(DetectionLayer, self).__init__()
+    self.anchors = anchors
+
+def create_modules(blocks):
+
+  net_info = blocks[0] # input and pre-processing
+  module_list = nn.ModuleList()
+  prev_filters = 3 # depth of last conv
+  output_filters = [] # number of output conv kernel，输出通道数量序列
+  
+  for index, x in enumerate(blocks[1:]):
+    module = nn.Sequential()
+    # convolutional模块有卷积层、批量归一化层和leaky ReLU激活层
+    if (x["type"] == "convolutional"):
+      # get layer info
+      activation = x["activation"]
+      try:
+        batch_normalize = int(x["batch_normalize"])
+        bias = False
+      except:
+        batch_normalize = 0
+        bias = True
+
+      filters = int(x["filters"]) # 卷积数量
+      padding = int(x["pad"]) # 填充数量
+      kernel_size = int(x["size"]) # 卷积核大小
+      stride= int(x["stride"]) # 步长
+
+      if padding:
+        padding = (kernel_size - 1) // 2 # 运算后，宽度和高度不变
+      else:
+        padding = 0
+      
+      # Add conv layer
+      conv = nn.Conv2d(prev_filters, filters, kernel_size, stride, padding, bias = bias)
+      module.add_module("conv_{0}".format(index), conv)
+
+      # Add batch norm layer
+      if batch_normalize:
+        bn = nn.BatchNorm2d(filters)
+        module.add_module("batch_norm_{0}".format(index), bn)
+      
+      # Check activation
+      if activation == "leaky":
+        activn = nn.LeakyReLU(0.1, inplace = True) # 斜率0.1
+        module.add_module("leaky_{0}".format(index), activn)
+
+    # upsample上采样层
+    elif (x["type"] == "upsample"):
+      stride = int(x["stride"])
+      upsample = nn.Upsample(scale_factor = 2, mode = "nearest") # 或者mode="bilinear"
+      module.add_module("upsample_module_list{}".format(index), upsample)
+
+    # route路由层，路由层是获取之前层的拼接
+    elif (x["type"] == "route"):
+      x["layers"] = x["layers"].split(",") # 保存start和end层号
+      # Start of a route
+      start = int(x["layers"][0])
+      # end, if there exists one
+      try:
+        end = int(x["layers"][1])
+      except:
+        end = 0 # 没有end
+      # Positive anotation
+      if start > 0:
+        start = start - index
+      if end > 0:
+        end = end - index
+      route = EmptyLayer() # 创建空层
+      module.add_module("route_{0}".format(index), route)
+      if end < 0:
+        # 计算卷积数量，即两层叠加
+        filters = output_filters[index + start] + output_filters[index + end]
+      else:
+        filters = output_filters[index + start] 
+
+    # shortcut捷径层（跳过连接），捷径层是将前一层的特征图添加到后面的层上
+    elif (x["type"] == "shortcut"):
+      shortcut = EmptyLayer()
+      module.add_module("shortcut_{}".format(index), shortcut)
+
+    # yolo层，检测层
+    elif (x["type"] == "yolo"):
+      # 保存mask序号
+      mask = x["mask"].split(",")
+      mask = [int(x) for x in mask]
+
+      # 保存anchors box
+      anchors = x["anchors"].split(",")
+      anchors = [int(a) for a in anchors]
+      # 两个一组，还ge和宽
+      anchors = [(anchors[i], anchors[i+1]) for i in range(0, len(anchors), 2)]
+      # 选取mask序号对应的anchors box，一般为3个
+      anchors = [anchors[i] for i in mask]
+      
+      detection = DetectionLayer(anchors)
+      module.add_module("Detection_{}".format(index), detection)
+    
+    module_list.append(module)
+    prev_filters = filters
+    output_filters.append(filters)
+
+  return (net_info, module_list)
+
+# 测试解析YOLO_v3配置文件
+# blocks = parse_cfg("cfg/yolov3.cfg")
+# print(create_modules(blocks))
+
+class Darknet(nn.Module):
+  # 用net_info和module_list对网络进行初始化
+  def __init__(self, cfgfile):
+    super(Darknet, self).__init__()
+    self.blocks = parse_cfg(cfgfile)
+    self.net_info, self.module_list = create_modules(self.blocks)
+
+  # CUDA为true，则用GPU加速前向传播
+  def forward(self, x, CUDA):
+
+    # delf.blocks第一个元素是net块
+    modules = self.blocks[1:]
+    # 缓存每个层的输出特征图，以备route层和shortcut层使用
+    outputs = {}
+
+    write = 0 # 是否遇到第一个检测图flag
+    for i, module in enumerate(modules):
+      module_type = (module["type"])
+
+      if module_type == "convolutional" or module_type == "upsample":
+        x = self.module_list[i](x)
+
+      elif module_type == "route":
+        layers = module["layers"]
+        layers = [int(a) for a in layers]
+
+        if layers[0] > 0:
+          layers[0] = layers[0] - i
+
+        if len(layers) == 1:
+          x = outputs[i + layers[0]]
+
+        else:
+          if layers[1] > 0:
+            layers[1] = layers[1] - i
+
+          map1 = outputs[i + layers[0]]
+          map2 = outputs[i + layers[1]]
+          x = torch.cat((map1, map2), 1) # 参数置1代表沿深度级联两个特征图
+
+      elif module_type == "shortcut":
+        from_ = int(module["from"])
+        x = outputs[i-1] + outputs[i+from_]
+
+      elif module_type == "yolo":
+        anchors = self.module_list[i][0].anchors
+        # input dimensions
+        inp_dim = int(self.net_info["height"])
+        
+        # number of classes
+        num_classes = int(module["classes"])
+
+        # transform
+        x = x.data
+        x = predict_transform(x, inp_dim, anchors, num_classes, CUDA)
+
+        if type(x) == int:
+          continue
+        # 如果收集器（容纳检测的张量）没有初始化
+        if not write:
+          detections = x
+          write = 1
+        else:
+          detections = torch.cat((detections, x), 1)
+      
+      outputs[i] = x
+
+    return detections
+
+# 测试向前传播
+# model = Darknet("cfg/yolov3.cfg")
+# inp = get_test_input()
+# pred = model(inp)
+# print(pred)
+# 张量形状1x10647x85，第一个维度是批量大小；85行，包括4个边界框属性(bx,by,bh,bw)、1个objectness分数和80个类别分数
+
+  def load_weights(self, weightfile):
+
+    """
+    权重属于归一化层和卷积层，权重存储顺序与配置文件层级顺序一致。
+    conv有shortcut，shortcut连接另一个conv，则先包含先前conv权重。
+    conv with batch norm：bn biases,bn weights,bn running_mean,bn running_var,conv weights
+    conv no batch norm：conv biases,conv weights
+    """
+    fp = open(weightfile, "rb")
+
+    # 标题信息：主版本，次版本，子版本，训练期间网络看到的图像
+    header = np.fromfile(fp, dtype = np.int32, count = 5)
+    self.header = torch.from_numpy(header)
+    self.seen = self.header[3]
+    
+    weights = np.fromfile(fp, dtype = np.float32)
+    # 迭代地加载权重文件到网络的模块上
+    ptr = 0 # 追踪权重数组位置指针
+    for i in range(len(self.module_list)):
+      module_type = self.blocks[i + 1]["type"] # 块包含第一块，模块不包含第一块
+
+      if module_type == "convolutional":
+        model = self.module_list[i]
+        # 根据conv模块是否有batch_normalize，加载权重
+        try:
+          batch_normalize = int(self.blocks[i+1]["batch_normalize"])
+        except:
+          batch_normalize = 0
+
+        conv = model[0]
+
+        # conv with batch norm
+        if (batch_normalize):
+          bn = model[1]
+          # 获取b_n layer权重的数量
+          num_bn_bias = bn.bias.numel()
+          # 加载权重
+          bn_bias = torch.from_numpy(weights[ptr:ptr+num_bn_bias])
+          ptr += num_bn_bias
+
+          bn_weight = torch.from_numpy(weights[ptr:ptr+num_bn_bias])
+          ptr += num_bn_bias
+
+          bn_running_mean = torch.from_numpy(weights[ptr:ptr+num_bn_bias])
+          ptr += num_bn_bias
+
+          bn_running_var = torch.from_numpy(weights[ptr:ptr+num_bn_bias])
+          ptr += num_bn_bias
+          # 根据模型权重的维度调整重塑加载的权重
+          bn_bias = bn_bias.view_as(bn.bias.data)
+          bn_weight = bn_weight.view_as(bn.weight.data)
+          bn_running_mean = bn_running_mean.view_as(bn.running_mean)
+          bn_running_var = bn_running_var.view_as(bn.running_var)
+          # 将数据复制到模型中
+          bn.bias.data.copy_(bn_bias)
+          bn.weight.data.copy_(bn_weight)
+          bn.running_mean.copy_(bn_running_mean)
+          bn.running_var.copy_(bn_running_var)
+
+        # conv no batch norm，只加载卷积层的偏置项
+        else:
+          # 偏置数量
+          num_bias = conv.bias.numel()
+          # 加载权重
+          conv_bias = torch.from_numpy(weights[ptr:ptr+num_bias])
+          ptr += num_bias
+          # 根据模型权重的维度调整重塑加载的权重
+          conv_bias = conv_bias.view_as(conv.bias.data)
+          # 将数据复制到模型中
+          conv.bias.data.copy_(conv_bias)
+        
+        # 最后，加载共有的卷积层权重
+        num_weight  = conv.weight.numel()
+        conv_weight = torch.from_numpy(weights[ptr:ptr+num_weight])
+        ptr += num_weight
+        conv_weight = conv_weight.view_as(conv.weight.data)
+        conv.weight.data.copy_(conv_weight)
+    
+# 测试加载预训练权重
+# model = Darknet("cfg/yolov3.cfg")
+# model.load_weights("yolov3.weights")
+# inp = get_test_input()
+# pred = model(inp)
+# print(pred)
+
+
+
+
+
+
+  

data/coco.names

@@ -0,0 +1,80 @@
+person
+bicycle
+car
+motorbike
+aeroplane
+bus
+train
+truck
+boat
+traffic light
+fire hydrant
+stop sign
+parking meter
+bench
+bird
+cat
+dog
+horse
+sheep
+cow
+elephant
+bear
+zebra
+giraffe
+backpack
+umbrella
+handbag
+tie
+suitcase
+frisbee
+skis
+snowboard
+sports ball
+kite
+baseball bat
+baseball glove
+skateboard
+surfboard
+tennis racket
+bottle
+wine glass
+cup
+fork
+knife
+spoon
+bowl
+banana
+apple
+sandwich
+orange
+broccoli
+carrot
+hot dog
+pizza
+donut
+cake
+chair
+sofa
+pottedplant
+bed
+diningtable
+toilet
+tvmonitor
+laptop
+mouse
+remote
+keyboard
+cell phone
+microwave
+oven
+toaster
+sink
+refrigerator
+book
+clock
+vase
+scissors
+teddy bear
+hair drier
+toothbrush

detect.py

@@ -0,0 +1,225 @@
+from __future__ import division
+
+import time
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+import numpy as np
+import cv2
+from util import *
+import argparse
+import os
+import os.path as osp
+from darknet import Darknet
+import pickle as pkl
+import pandas as pd
+import random
+
+# 命令行参数
+def arg_parse():
+   
+    parser = argparse.ArgumentParser(description='YOLO v3 Detection Module')
+
+    # images（用于指定输入图像或图像目录）
+    parser.add_argument("--images", dest = 'images', help = 
+                        "Image / Directory containing images to perform detection upon",
+                        default = "imgs", type = str)
+    # det（保存检测结果的目录）
+    parser.add_argument("--det", dest = 'det', help = 
+                        "Image / Directory to store detections to",
+                        default = "det", type = str)
+    # batch大小
+    parser.add_argument("--bs", dest = "bs", help = "Batch size", default = 1)
+    # objectness置信度
+    parser.add_argument("--confidence", dest = "confidence", help = "Object Confidence to filter predictions", default = 0.5)
+    # NMS阈值
+    parser.add_argument("--nms_thresh", dest = "nms_thresh", help = "NMS Threshhold", default = 0.4)
+    # cfg（替代配置文件）
+    parser.add_argument("--cfg", dest = 'cfgfile', help = 
+                        "Config file",
+                        default = "cfg/yolov3.cfg", type = str)
+    parser.add_argument("--weights", dest = 'weightsfile', help = 
+                        "weightsfile",
+                        default = "yolov3.weights", type = str)
+    # reso（输入图像的分辨率，可用于在速度与准确度之间的权衡）
+    parser.add_argument("--reso", dest = 'reso', help = 
+                        "Input resolution of the network. Increase to increase accuracy. Decrease to increase speed",
+                        default = "416", type = str)
+
+    return parser.parse_args()
+
+if __name__ == '__main__':
+
+    args = arg_parse()
+    images = args.images
+    batch_size = int(args.bs)
+    confidence = float(args.confidence)
+    nms_thesh = float(args.nms_thresh)
+    start = 0
+    CUDA = torch.cuda.is_available()
+    num_classes = 80 # COCO数据集中目标的名称
+    classes = load_classes("data/coco.names")
+        
+    # 初始化网络，加载权重
+    print("正在加载网络QAQ")
+    model = Darknet(args.cfgfile)
+    model.load_weights(args.weightsfile)
+    print("网络加载成功QvQ")
+
+    model.net_info["height"] = args.reso
+    inp_dim = int(model.net_info["height"])
+    assert inp_dim % 32 == 0
+    assert inp_dim > 32
+
+    # GPU加速
+    if CUDA:
+        model.cuda()
+
+    # 模型评估
+    model.eval()
+
+    # 从磁盘读取图像或从目录读取多张图像，图像路径imlist
+    read_dir = time.time() # 测量时间的检查点
+    # 检测阶段
+    try:
+        imlist = [osp.join(osp.realpath('.'), images, img) for img in os.listdir(images)]
+    except NotADirectoryError:
+        imlist = []
+        imlist.append(osp.join(osp.realpath('.'), images))
+    except FileNotFoundError:
+        print("没有找到{}文件或目录QwQ".format(images))
+        exit()
+
+    # 如果没有保存检测结果的目录，就创建一个
+    if not os.path.exists(args.det):
+        os.makedirs(args.det)
+
+    # 用OpenCV加载多张图片图像
+    load_batch = time.time()
+    loaded_ims = [cv2.imread(x) for x in imlist]
+    # 转成PyTorch图像格式
+    im_batches = list(map(prep_image, loaded_ims, [inp_dim for x in range(len(imlist))]))
+    # 包含原始图像的维度的列表
+    im_dim_list = [(x.shape[1], x.shape[0]) for x in loaded_ims]
+    im_dim_list = torch.FloatTensor(im_dim_list).repeat(1,2)
+        
+    # 创建batch
+    leftover = 0
+    if (len(im_dim_list) % batch_size):
+        leftover = 1
+    if batch_size != 1:
+        num_batches = len(imlist) // batch_size + leftover
+        im_batches = [torch.cat((im_batches[i*batch_size : min((i+1)*batch_size,
+                    len(im_batches))])) for i in range(num_batches)]
+
+    write = 0
+
+    if CUDA:
+        im_dim_list = im_dim_list.cuda()
+            
+    start_det_loop = time.time()
+    for i, batch in enumerate(im_batches):
+        # 载入图片
+        start = time.time()
+        if CUDA:
+            batch = batch.cuda()
+        with torch.no_grad():
+            prediction = model(Variable(batch), CUDA)
+        prediction = write_results(prediction, confidence, num_classes, nms_conf=nms_thesh)
+            
+        end = time.time()
+
+        if type(prediction) == int:
+            for im_num, image in enumerate(imlist[i*batch_size: min((i +  1)*batch_size, len(imlist))]):
+                im_id = i*batch_size + im_num
+                print("{0:20s} 预测用时{1:6.3f} 秒".format(image.split("/")[-1], (end - start)/batch_size))
+                print("{0:20s} {1:s}".format("检测到的对象：", " "))
+                print("----------------------------------------------------------")
+            continue
+
+        prediction[:,0] += i*batch_size # 将batch索引转换成imlist索引
+        
+        if not write: # 初始化output
+            output = prediction
+            write = 1
+        else:
+            output = torch.cat((output, prediction))
+            
+        for im_num, image in enumerate(imlist[i*batch_size:min((i+1)*batch_size, len(imlist))]):
+            im_id = i*batch_size + im_num
+            objs = [classes[int(x[-1])] for x in output if int(x[0]) == im_id]
+            print("{0:20s} 预测用时{1:6.3f} 秒".format(image.split("/")[-1], (end - start)/batch_size))
+            print("{0:20s} {1:s}".format("检测到的对象:", " ".join(objs)))
+            print("----------------------------------------------------------------")
+
+        if CUDA:
+            torch.cuda.synchronize() # 保证CUDA核与CPU同步
+
+    # 在图像上绘制边界框
+    try:
+        output
+    except NameError:
+        print("不存在检测结果TAT")
+        exit()
+        
+    # 输出边界框对应网络输入大小，需要将边界框属性转换到图像的原始尺寸
+    im_dim_list = torch.index_select(im_dim_list, 0, output[:,0].long())
+    scaling_factor = torch.min(inp_dim/im_dim_list,1)[0].view(-1,1)
+        
+    output[:,[1,3]] -= (inp_dim - scaling_factor*im_dim_list[:,0].view(-1,1))/2
+    output[:,[2,4]] -= (inp_dim - scaling_factor*im_dim_list[:,1].view(-1,1))/2
+    output[:,1:5] /= scaling_factor
+        
+    for i in range(output.shape[0]):
+        output[i, [1,3]] = torch.clamp(output[i, [1,3]], 0.0, im_dim_list[i,0])
+        output[i, [2,4]] = torch.clamp(output[i, [2,4]], 0.0, im_dim_list[i,1])
+            
+    output_recast = time.time()
+    
+    # 随机选择的颜色用于绘制边界框
+    class_load = time.time()
+    colors = pkl.load(open("pallete", "rb"))
+
+    #　开始绘制边界框
+    draw = time.time()
+
+    # 绘制边界框:从colors中随机选颜色绘制矩形框
+    # 边界框左上角创建一个填充后的矩形，写入该框位置检测到的目标的类别
+    def write(x, results):
+
+        c1 = tuple(x[1:3].int())
+        c2 = tuple(x[3:5].int())
+        img = results[int(x[0])]
+        cls = int(x[-1])
+        color = random.choice(colors)
+        label = "{0}".format(classes[cls])
+        cv2.rectangle(img, c1, c2, color, 1)
+        t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_PLAIN, 1, 1)[0]
+        c2 = c1[0] + t_size[0] + 3, c1[1] + t_size[1] + 4
+        cv2.rectangle(img, c1, c2, color, -1) # -1表示填充的矩形
+        cv2.putText(img, label, (c1[0], c1[1] + t_size[1] + 4), cv2.FONT_HERSHEY_PLAIN, 1, [225,225,225], 1)
+
+        return img
+        
+    list(map(lambda x:write(x, loaded_ims), output))
+
+    # 保存检测结果图像,det_图像名
+    det_names = pd.Series(imlist).apply(lambda x: "{}/det_{}".format(args.det, x.split("/")[-1]))
+    # 将带有检测结果的图像写入det_names中的地址
+    list(map(cv2.imwrite, det_names, loaded_ims))
+    end = time.time()
+
+    # 显示输出时间的总结
+    print("总结")
+    print("----------------------------------------------------------------")
+    print("{:25s} {}".format("任务", "所用时间(s)"))
+    print()
+    print("{:25s} {:2.3f}".format("读入目录", load_batch - read_dir))
+    print("{:25s} {:2.3f}".format("加载batch", start_det_loop - load_batch))
+    print("{:25s} {:2.3f}".format("检测(" + str(len(imlist)) + "张图)", output_recast - start_det_loop))
+    print("{:25s} {:2.3f}".format("输出处理", class_load - output_recast))
+    print("{:25s} {:2.3f}".format("绘制边界框", end - draw))
+    print("{:25s} {:2.3f}".format("平均检测时间", (end - load_batch)/len(imlist)))
+    print("----------------------------------------------------------------")
+
+    torch.cuda.empty_cache()

util.py

@@ -0,0 +1,229 @@
+from __future__ import division
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.autograd import Variable
+import numpy as np
+import cv2
+
+# 获取任意给定图像中存在的类别
+def unique(tensor):
+    tensor_np =tensor.cpu().numpy()
+    unique_np = np.unique(tensor_np)
+    unique_tensor = torch.from_numpy(unique_np)
+
+    tensor_res = tensor.new(unique_tensor.shape)
+    tensor_res.copy_(unique_tensor)
+
+    return tensor_res
+
+# 计算两个边界框的IoU
+def bbox_iou(box1, box2):
+    # 获取边框的坐标
+    b1_x1, b1_y1, b1_x2, b1_y2 = box1[:,0], box1[:,1], box1[:,2], box1[:,3]
+    b2_x1, b2_y1, b2_x2, b2_y2 = box2[:,0], box2[:,1], box2[:,2], box2[:,3]
+
+    # 获取交叉矩形的坐标
+    inter_rect_x1 = torch.max(b1_x1, b2_x1)
+    inter_rect_y1 = torch.max(b1_y1, b2_y1)
+    inter_rect_x2 = torch.min(b1_x2, b2_x2)
+    inter_rect_y2 = torch.min(b1_y2, b2_y2)
+
+    # 交叉面积
+    inter_area = torch.clamp(inter_rect_x2 - inter_rect_x1 + 1, min=0) * torch.clamp(inter_rect_y2 - inter_rect_y1 + 1, min=0)
+   
+    # 合并面积
+    b1_area = (b1_x2 - b1_x1 + 1) * (b1_y2 - b1_y1 + 1)
+    b2_area = (b2_x2 - b2_x1 + 1) * (b2_y2 - b2_y1 + 1)
+    union_area = b1_area + b2_area - inter_area
+
+    # IoU
+    iou = inter_area / union_area
+
+    return iou
+
+# 把检测特征图转换成二维张量，张量的每一行对应边界框的属性，5个参数：输出，输入图像的维度……
+def predict_transform(prediction, inp_dim, anchors, num_classes, CUDA=True):
+    
+    batch_size = prediction.size(0)
+    stride = inp_dim // prediction.size(2)
+    grid_size = inp_dim // stride
+    bbox_attrs = 5 + num_classes
+    num_anchors = len(anchors)
+
+    prediction = prediction.view(batch_size, bbox_attrs*num_anchors, grid_size*grid_size)
+    prediction = prediction.transpose(1,2).contiguous()
+    prediction = prediction.view(batch_size, grid_size*grid_size*num_anchors, bbox_attrs)
+    
+    # 锚点的维度与net块的h和w属性一致，输入图像的维度和检测图的维度之商就是步长，用检测特征图的步长分割锚点
+    anchors = [(a[0]/stride, a[1]/stride) for a in anchors]
+
+    # 对（x，y）坐标和objectness分数执行Sigmoid函数操作
+    prediction[:,:,0] = torch.sigmoid(prediction[:,:,0])
+    prediction[:,:,1] = torch.sigmoid(prediction[:,:,1])
+    prediction[:,:,4] = torch.sigmoid(prediction[:,:,4])
+
+    # 将网格偏移添加到中心坐标预测中
+    grid = np.arange(grid_size)
+    a,b = np.meshgrid(grid, grid)
+
+    x_offset = torch.FloatTensor(a).view(-1,1)
+    y_offset = torch.FloatTensor(b).view(-1,1)
+
+    if CUDA:
+        x_offset = x_offset.cuda()
+        y_offset = y_offset.cuda()
+    
+    x_y_offset = torch.cat((x_offset, y_offset), 1).repeat(1,num_anchors).view(-1,2).unsqueeze(0)
+
+    prediction[:,:,:2] += x_y_offset
+
+    # 将锚点应用到边界框维度中
+    anchors = torch.FloatTensor(anchors)
+
+    if CUDA:
+        anchors = anchors.cuda()
+    
+    anchors = anchors.repeat(grid_size*grid_size, 1).unsqueeze(0)
+    prediction[:,:,2:4] = torch.exp(prediction[:,:,2:4])*anchors
+
+    # 将sigmoid激活函数应用到类别分数中
+    prediction[:,:,5:5 + num_classes] = torch.sigmoid((prediction[:,:,5:5 + num_classes]))
+
+    # 将检测图的大小调整到与输入图像大小一致，乘以stride变量（边界框属性根据特征图大小而定）
+    prediction[:,:,:4] *= stride
+
+    return prediction
+
+# 加载类别，返回字典——将每个类别的索引映射到其名称的字符串
+def load_classes(namesfile):
+    fp = open(namesfile, "r")
+    names = fp.read().split("\n")[:-1]
+    return names
+
+# 输出满足objectness分数阈值和非极大值抑制(NMS)，得到真实检测结果
+def write_results(prediction, confidence, num_classes, nms_conf=0.4):
+    # 输入为预测结果，置信度，类别数，NMS阈值
+
+    # 低于objectness分数的每个边界框，其每个属性值都置0，即一整行。
+    conf_mask = (prediction[:,:,4] > confidence).float().unsqueeze(2)
+    prediction = prediction*conf_mask
+
+    # 每个框的两个对焦坐标更容易计算两个框的IoU，故将(中心x，中心y，高度，宽度)属性转化成(左上角x，左上角y，右下角x，右下角y)
+    box_a = prediction.new(prediction.shape)
+    box_a[:,:,0] = (prediction[:,:,0] - prediction[:,:,2]/2)
+    box_a[:,:,1] = (prediction[:,:,1] - prediction[:,:,3]/2)
+    box_a[:,:,2] = (prediction[:,:,0] + prediction[:,:,2]/2)
+    box_a[:,:,3] = (prediction[:,:,1] + prediction[:,:,3]/2)
+    prediction[:,:,:4] = box_a[:,:,:4]
+
+    batch_size = prediction.size(0)
+
+    #output = prediction.new(1, prediction.size(2) + 1)
+    write = False # 标识尚未初始化输出
+
+    # 在第一个维度即bacth上循环，一次完成一个图像的置信度阈值和NMS
+    for ind in range(batch_size):
+        # 获取图像,10647x85
+        image_pred = prediction[ind]
+
+        # 每个边界框行有85个属性，其中80个类别分数，只取最大值的类别分数
+        # 获取具有最高分数的类及其索引
+        max_conf, max_conf_score = torch.max(image_pred[:,5:5+num_classes], 1)
+        max_conf = max_conf.float().unsqueeze(1)
+        max_conf_score = max_conf_score.float().unsqueeze(1)
+        # 删除80个分类分数，增加最高分数类别的索引及最高分数
+        seq = (image_pred[:,:5], max_conf, max_conf_score)
+        image_pred = torch.cat(seq, 1)
+
+        # 删除objectness置信度小于阈值的置0条目,try-except处理无检测结果的情况，continue跳过对本图像的循环
+        non_zero_ind = torch.nonzero(image_pred[:,4])
+        try:
+            image_pred_ = image_pred[non_zero_ind.squeeze(),:].view(-1,7) # 7列
+        except:
+            continue
+        
+        # PyTorch 0.4兼容
+        if image_pred_.shape[0] == 0:
+            continue
+
+        # 获得一个图像的所有种类
+        img_classes = unique(image_pred_[:,-1])
+        
+        # 按类别执行NMS
+        for cls in img_classes:
+            # 得到一个类别的所有检测
+            cls_mask = image_pred_*(image_pred_[:,-1] == cls).float().unsqueeze(1) 
+            class_mask_ind = torch.nonzero(cls_mask[:,-2]).squeeze()
+
+            image_pred_class = image_pred_[class_mask_ind].view(-1,7)
+
+            # 对所有检测排序,按照objectness置信度
+            conf_sort_index = torch.sort(image_pred_class[:,4], descending=True)[1]
+            image_pred_class = image_pred_class[conf_sort_index]
+            idx = image_pred_class.size(0)
+
+            # 对于每一个检测,执行NMS
+            for i in range(idx):
+                # 获取正在查看的box之后所有boxes的IoUs
+                try:
+                    ious = bbox_iou(image_pred_class[i].unsqueeze(0), image_pred_class[i+1:])
+                except ValueError: # image_pred_class[i+1,:]返回空张量
+                    break
+                except IndexError: # image_pred_class移除部分后，idx索引越界
+                    break
+                    
+                # 清除IoU>阈值的检测
+                iou_mask = (ious < nms_conf).float().unsqueeze(1)
+                image_pred_class[i+1:] *= iou_mask
+
+                # 移除0条目
+                non_zero_ind = torch.nonzero(image_pred_class[:,4]).squeeze()
+                image_pred_class = image_pred_class[non_zero_ind].view(-1,7)
+
+            batch_ind = image_pred_class.new(image_pred_class.size(0), 1).fill_(ind)
+            seq = batch_ind, image_pred_class
+
+            if not write:
+                output = torch.cat(seq, 1)
+                write = True
+            else:
+                out = torch.cat(seq, 1)
+                output = torch.cat((output, out))
+    # 输出一个形状为Dx8的张量；其中D是所有图像中的「真实」检测结果，每个都用一行表示。
+    # 每一个检测结果都有8个属性，即该检测结果所属的batch中图像的索引、4个对角的坐标、objectness分数、有最大置信度的类别的分数、该类别的索引。
+    try:
+        return output
+    except:
+        return 0
+
+# 使用填充调整具有不变长宽性的图像
+def letterbox_image(img, inp_dim):
+    
+    img_w, img_h = img.shape[1], img.shape[0]
+    w, h = inp_dim
+    new_w = int(img_w * min(w/img_w, h/img_h))
+    new_h = int(img_h * min(w/img_w, h/img_h))
+    resized_image = cv2.resize(img, (new_w,new_h), interpolation = cv2.INTER_CUBIC)
+    
+    canvas = np.full((inp_dim[1], inp_dim[0], 3), 128)
+
+    canvas[(h-new_h)//2:(h-new_h)//2 + new_h,(w-new_w)//2:(w-new_w)//2 + new_w,  :] = resized_image
+    
+    return canvas
+
+# 将numpy数组转换成PyTorch的的输入格式
+# OpenCV将图像载入成numpy数组，颜色通道为BGR。
+# PyTorch的图像输入格式是(batch x 通道 x 高度 x 宽度)，通道顺序RGB。
+def prep_image(img, inp_dim):
+
+    img = letterbox_image(img, (inp_dim, inp_dim)) # 转换格式大小
+    img = img[:,:,::-1].transpose((2,0,1)).copy() # BGR -> RGB(起止位置省略，步长为-1，负:从右往左)) | H x W x C -> C x H x W
+    img = torch.from_numpy(img).float().div(255.0).unsqueeze(0)
+
+    return img
+        
+
+
+    

video.py

@@ -0,0 +1,157 @@
+# 在视频/网络摄像头上运行检测器
+# 不在batch上迭代，而是在视频的帧上迭代
+
+from __future__ import division
+
+import time
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+import numpy as np
+import cv2
+from util import *
+import argparse
+import os
+import os.path as osp
+from darknet import Darknet
+import pickle as pkl
+import pandas as pd
+import random
+
+# 命令行参数
+def arg_parse():
+   
+    parser = argparse.ArgumentParser(description='YOLO v3 Detection Module')
+
+    # images（用于指定输入图像或图像目录）
+    parser.add_argument("--images", dest = 'images', help = 
+                        "Image / Directory containing images to perform detection upon",
+                        default = "imgs", type = str)
+    # det（保存检测结果的目录）
+    parser.add_argument("--det", dest = 'det', help = 
+                        "Image / Directory to store detections to",
+                        default = "det", type = str)
+    # batch大小
+    parser.add_argument("--bs", dest = "bs", help = "Batch size", default = 1)
+    # objectness置信度
+    parser.add_argument("--confidence", dest = "confidence", help = "Object Confidence to filter predictions", default = 0.5)
+    # NMS阈值
+    parser.add_argument("--nms_thresh", dest = "nms_thresh", help = "NMS Threshhold", default = 0.4)
+    # cfg（替代配置文件）
+    parser.add_argument("--cfg", dest = 'cfgfile', help = 
+                        "Config file",
+                        default = "cfg/yolov3.cfg", type = str)
+    parser.add_argument("--weights", dest = 'weightsfile', help = 
+                        "weightsfile",
+                        default = "yolov3.weights", type = str)
+    # reso（输入图像的分辨率，可用于在速度与准确度之间的权衡）
+    parser.add_argument("--reso", dest = 'reso', help = 
+                        "Input resolution of the network. Increase to increase accuracy. Decrease to increase speed",
+                        default = "416", type = str)
+
+    return parser.parse_args()
+
+if __name__ == '__main__':
+
+    args = arg_parse()
+    images = args.images
+    batch_size = int(args.bs)
+    confidence = float(args.confidence)
+    nms_thesh = float(args.nms_thresh)
+    start = 0
+    CUDA = torch.cuda.is_available()
+    num_classes = 80 # COCO数据集中目标的名称
+    classes = load_classes("data/coco.names")
+        
+    # 初始化网络，加载权重
+    print("正在加载网络QAQ")
+    model = Darknet(args.cfgfile)
+    model.load_weights(args.weightsfile)
+    print("网络加载成功QvQ")
+
+    model.net_info["height"] = args.reso
+    inp_dim = int(model.net_info["height"])
+    assert inp_dim % 32 == 0
+    assert inp_dim > 32
+
+    # GPU加速
+    if CUDA:
+        model.cuda()
+    
+    # 模型评估
+    model.eval()
+
+    # 绘制边界框:从colors中随机选颜色绘制矩形框
+    # 边界框左上角创建一个填充后的矩形，写入该框位置检测到的目标的类别
+    def write(x, results):
+
+        c1 = tuple(x[1:3].int())
+        c2 = tuple(x[3:5].int())
+        img = results # 仅处理一帧
+        cls = int(x[-1])
+        color = random.choice(colors)
+        label = "{0}".format(classes[cls])
+        cv2.rectangle(img, c1, c2, color, 1)
+        t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_PLAIN, 1, 1)[0]
+        c2 = c1[0] + t_size[0] + 3, c1[1] + t_size[1] + 4
+        cv2.rectangle(img, c1, c2, color, -1) # -1表示填充的矩形
+        cv2.putText(img, label, (c1[0], c1[1] + t_size[1] + 4), cv2.FONT_HERSHEY_PLAIN, 1, [225,225,225], 1)
+
+        return img
+
+    # 检测阶段
+    videofile = "workingcell.mp4" # 加载视频文件路径
+    cap = cv2.VideoCapture(videofile) # 用OpenCV打开视频/相机流
+    #assert cap.isOpened(), '未找到需要检测视频TAT'
+    
+    frames = 0 # 帧的数量
+    start = time.time()
+    # 在帧上迭代，一次处理一帧
+    while cap.isOpened():
+        ret, frame = cap.read()
+
+        if ret:
+            img = prep_image(frame, inp_dim)
+            im_dim = frame.shape[1], frame.shape[0]
+            im_dim = torch.FloatTensor(im_dim).repeat(1,2)
+
+            if CUDA:
+                im_dim = im_dim.cuda()
+                img = img.cuda()
+            
+            output = model(Variable(img, volatile=True), CUDA)
+            output = write_results(output, confidence, num_classes, nms_conf=nms_thesh)
+
+            if type(output) == int:
+                frames += 1
+                print("视频的FPS为 {:5.4f}".format(frames / (time.time() - start)))
+                # 使用cv2.imshow展示画有边界框的帧
+                cv2.imshow("帧", frame)
+                key = cv2.waitKey(1)
+                # 用户按q，就会终止视频(代码中断循环)
+                if key & 0xFF == ord('q'):
+                    break
+                continue
+
+            output[:,1:5] = torch.clamp(output[:,1:5], 0.0, float(inp_dim))
+            im_dim = im_dim.repeat(output.size(0), 1)/inp_dim
+            output[:,1:5] *= im_dim
+
+            classes = load_classes('data/coco.names')
+            colors = pkl.load(open("pallete", "rb"))
+            list(map(lambda x: write(x, frame), output))
+
+            cv2.imshow("帧", frame)
+            key = cv2.waitKey(1)
+            if key & 0xFF == ord('q'):
+                    break
+            frames += 1
+            print(time.time() - start)
+            print("视频的FPS为 {:5.4f}".format(frames / (time.time() - start)))
+        
+        else:
+            break
+
+
+
+