Win10 tensorflow Faster-RCNN VOC2007+Ubuntu16.04 YOLOV3

老师让测试一下几个检测模型，Faster-RCNN caffe版本编译一直有问题，Github上找了一个Windows下tensorflow的版本，全部默认参数没有调整，时间比较紧，简单跑了一下，显卡GTX1080 8G测试结果如下。
Faster-RCNN，使用VGG16预训练模型，基于windows10，tensorflow，cuda9.0,cudnn7.1,Python3.5，VOC2007数据集约10000张图片，类别20,训练集测试集1:1，迭代40000次，约7小时。
AP for aeroplane = 0.6837
AP for bicycle = 0.7048
AP for bird = 0.5932
AP for boat = 0.5433
AP for bottle = 0.3847
AP for bus = 0.6699
AP for car = 0.7083
AP for cat = 0.7078
AP for chair = 0.3824
AP for cow = 0.7424
AP for diningtable = 0.6268
AP for dog = 0.7494
AP for horse = 0.7963
AP for motorbike = 0.6616
AP for person = 0.6972
AP for pottedplant = 0.3255
AP for sheep = 0.6343
AP for sofa = 0.6185
AP for train = 0.6921
AP for tvmonitor = 0.5006
Mean AP = 0.6211

YOLOv3,batch=48,subdivisions=16,初始学习率0.001,使用darknet53.conv.74预训练模型，基于ubuntu16.04,darknet,cuda8.0,cudnn5.1,VOC2007数据集约10000张图片，类别20,训练集测试集1:1，迭代10000轮，约15小时。
AP for aeroplane = 0.656469218822
AP for bicycle = 0.738171052772
AP for bird = 0.599342814029
AP for boat = 0.466246285632
AP for bottle = 0.377019995151
AP for bus = 0.776727860512
AP for car = 0.81492015551
AP for cat = 0.769212772823
AP for chair = 0.443830951069
AP for cow = 0.713351960799
AP for diningtable = 0.519093086527
AP for dog = 0.719592841968
AP for horse = 0.788485138533
AP for motorbike = 0.790403446572
AP for person = 0.737553938525
AP for pottedplant = 0.311726030682
AP for sheep = 0.613778402001
AP for sofa = 0.661944290905
AP for train = 0.724236888543
AP for tvmonitor = 0.624288643631
Mean AP = 0.64231978875

其中Faster-RCNN计算mAP，参考test_net

# --------------------------------------------------------
# Tensorflow Faster R-CNN
# Licensed under The MIT License [see LICENSE for details]
# Written by Zheqi he, Xinlei Chen, based on code from Ross Girshick
# --------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

# import _init_paths
from lib.utils.test import test_net
from lib.config import config as cfg
# from lib.config import cfg, cfg_from_file, cfg_from_list
from lib.datasets.factory import get_imdb
import argparse
import pprint
import time, os, sys

import tensorflow as tf
from lib.nets.vgg16 import vgg16
# from nets.resnet_v1 import resnetv1
# from nets.mobilenet_v1 import mobilenetv1

demonet = 'vgg16'
dataset = 'pascal_voc'
NETS = {'vgg16': ('vgg16.ckpt',)}
DATASETS = {'pascal_voc': ('voc_2007_trainval',)}
tfmodel = os.path.join('output', demonet, DATASETS[dataset][0], 'default', NETS[demonet][0])

def parse_args():
  """
  Parse input arguments
  """
  parser = argparse.ArgumentParser(description='Test a Fast R-CNN network')
  parser.add_argument('--cfg', dest='cfg_file',
            help='optional config file', default=None, type=str)
  parser.add_argument('--model', dest='model',
            help='model to test',
            default=None, type=str)
  parser.add_argument('--imdb', dest='imdb_name',
            help='dataset to test',
            default='voc_2007_test', type=str)
  parser.add_argument('--comp', dest='comp_mode', help='competition mode',
            action='store_true')
  parser.add_argument('--num_dets', dest='max_per_image',
            help='max number of detections per image',
            default=100, type=int)
  parser.add_argument('--tag', dest='tag',
                        help='tag of the model',
                        default='', type=str)
  # parser.add_argument('--net', dest='net',
  #                     help='vgg16, res50, res101, res152, mobile',
  #                     default='res50', type=str)
  parser.add_argument('--set', dest='set_cfgs',
                        help='set config keys', default=None,
                        nargs=argparse.REMAINDER)
  parser.add_argument('--net', dest='demo_net', help='Network to use [vgg16 res101]',
                      choices=NETS.keys(), default='vgg16')
  parser.add_argument('--dataset', dest='dataset', help='Trained dataset [pascal_voc pascal_voc_0712]',
                      choices=DATASETS.keys(), default='pascal_voc_0712')

  # if len(sys.argv) == 1:
  #   parser.print_help()
  #   sys.exit(1)

  args = parser.parse_args()                 # 解析参数
  return args

if __name__ == '__main__':
  args = parse_args()

  print('Called with args:')
  print(args)

  # if args.cfg_file is not None:
  #   cfg_from_file(args.cfg_file)
  # if args.set_cfgs is not None:
  #   cfg_from_list(args.set_cfgs)

  print('Using config:')
  pprint.pprint(cfg)

  # if has model, get the name from it
  # if does not, then just use the initialization weights
  if tfmodel:
    filename = os.path.splitext(os.path.basename(tfmodel))[0]
  else:
    filename = os.path.splitext(os.path.basename(args.weight))[0]

  tag = args.tag
  tag = tag if tag else 'default'
  filename = tag + '/' + filename

  imdb = get_imdb(args.imdb_name)
  imdb.competition_mode(args.comp_mode)

  tfconfig = tf.ConfigProto(allow_soft_placement=True)
  tfconfig.gpu_options.allow_growth=True

  # init session
  sess = tf.Session(config=tfconfig)
  # load network



  if demonet == 'vgg16':
      net = vgg16()
  # elif args.net == 'res50':
  #   net = resnetv1(num_layers=50)
  # elif args.net == 'res101':
  #   net = resnetv1(num_layers=101)
  # elif args.net == 'res152':
  #   net = resnetv1(num_layers=152)
  # elif args.net == 'mobile':
  #   net = mobilenetv1()
  else:
    raise NotImplementedError

  # load model
  net.create_architecture(sess, mode="TEST",num_classes=6, tag='default',
                          anchor_scales=[8, 16, 32],
                          anchor_ratios=[0.5, 1, 2])

  if tfmodel:
    print(('Loading model check point from {:s}').format(tfmodel))
    saver = tf.train.Saver()
    saver.restore(sess, tfmodel)
    print('Loaded.')
  else:
    print(('Loading initial weights from {:s}').format(args.weight))
    sess.run(tf.global_variables_initializer())
    print('Loaded.')

  test_net(sess, net, imdb, filename, max_per_image=args.max_per_image)

  sess.close()

其中记得修改num_classes=6为自己的类别数量，例如我的VOC数据集为20种+背景，num_classes=21。