faster-rcnn 笔记

2019-02-18,15点00

'''
下面是别人写的原始的笔记,我在上面自己补充了一些.

'''
#https://www.cnblogs.com/the-home-of-123/p/9747963.html

 #  以voc数据集为例，按照imdb的命名，利用pascal_voc()函数生成不同的imdb

'''

for year in ['2007', '2012']:
  for split in ['train', 'val', 'trainval', 'test']:
    name = 'voc_{}_{}'.format(year, split)  #year='2007', split='trainval'
    __sets[name] = (lambda split=split, year=year: pascal_voc(split, year))

def get_imdb(name):
  """Get an imdb (image database) by name."""
  if name not in __sets:
    raise KeyError('Unknown dataset: {}'.format(name))
  return __sets[name]()

'''

# self._data_path = os.path.join(self._devkit_path, 'VOC' + self._year)  #数据库路径
# self._classes = ('__background__',  # always index 0， 训练类别标签，包含背景类
#                   'person')
#  # Default to roidb handler
# self._roidb_handler = self.gt_roidb #感兴趣区域（ROI）数据库
# self._salt = str(uuid.uuid4()) #？？
# self._comp_id = 'comp4' # ？？

def _build_network(self, is_training=True):
    # select initializers进行初始化
    if cfg.TRAIN.TRUNCATED:
      initializer = tf.truncated_normal_initializer(mean=0.0, stddev=0.01)
      initializer_bbox = tf.truncated_normal_initializer(mean=0.0, stddev=0.001)
    else:
      initializer = tf.random_normal_initializer(mean=0.0, stddev=0.01)
      initializer_bbox = tf.random_normal_initializer(mean=0.0, stddev=0.001)

    net_conv = self._image_to_head(is_training)##经过特征提取网络，初步提取特征
    with tf.variable_scope(self._scope, self._scope):
      # build the anchors for the image
      self._anchor_component()###产生anchor
      # region proposal network ###产生proposal的坐标
      rois = self._region_proposal(net_conv, is_training, initializer)
      #这里面rois表示的是那些非背景的区域对应到feature_map上的坐标组成的数组.
      '''
      上面一行的代码是和兴!!
      '''
      # region of interest pooling
      if cfg.POOLING_MODE == 'crop':
        pool5 = self._crop_pool_layer(net_conv, rois, "pool5") ###对产生的porposal进行ROI池化，统一格式
      else:
        raise NotImplementedError
      '''
      这里面得到的pool5就是把rois
      '''

    fc7 = self._head_to_tail(pool5, is_training)
    with tf.variable_scope(self._scope, self._scope):
      # region classification 输入到Fast-RCNN网络中，对样本进行分类和预测框回归
      cls_prob, bbox_pred = self._region_classification(fc7, is_training,
                                                        initializer, initializer_bbox)

    '''
    利用self._region_classification 里面的fc 层和softmax层输出 cls_prob, bbox_pred.得到最总的预测结果.
    '''
    self._score_summaries.update(self._predictions)

    return rois, cls_prob, bbox_pred

'''
下面是上面说的核心代码的分析
'''

def _region_proposal(self, net_conv, is_training, initializer):
    rpn = slim.conv2d(net_conv, cfg.RPN_CHANNELS, [3, 3], trainable=is_training, weights_initializer=initializer,
                        scope="rpn_conv/3x3") ##经过一个3X3卷积，之后分两条线
    self._act_summaries.append(rpn)

    '''
    下面一行的代码就是唐老师说的最精髓的地方.
    anchors这些概念都是虚拟的.其实都没有.都是通过学习得到的.

    输出的维度是self._num_anchors * 2,   每2个数表示一种anchor对应的得分.至于到底哪个数对应哪个anchor,
    不用指明,这些完全是通过学习获得的.这样避免人工干预,效果更好.更加end_to_end.

    从这里面kernal=[1*1]就表示每一个像素点对应9个anchor!

    '''

    rpn_cls_score = slim.conv2d(rpn, self._num_anchors * 2, [1, 1], trainable=is_training,
                                weights_initializer=initializer,
                                padding='VALID', activation_fn=None, scope='rpn_cls_score') ###第一条线产生预测类别确定是背景还是类别

    '''
    我纳闷的地方是这里面得到的rpn_cls_score:(1,height,width,18) 表示的是9个框的分数.而表示不了各个分类的分数
    那么后面的nms怎么做?

    其实这个地方只是对是否是背景做nms
    '''

    # change it so that the score has 2 as its channel size
    rpn_cls_score_reshape = self._reshape_layer(rpn_cls_score, 2, 'rpn_cls_score_reshape')
    rpn_cls_prob_reshape = self._softmax_layer(rpn_cls_score_reshape, "rpn_cls_prob_reshape")
    rpn_cls_pred = tf.argmax(tf.reshape(rpn_cls_score_reshape, [-1, 2]), axis=1, name="rpn_cls_pred")

    rpn_cls_prob = self._reshape_layer(rpn_cls_prob_reshape, self._num_anchors * 2, "rpn_cls_prob")

    rpn_bbox_pred = slim.conv2d(rpn, self._num_anchors * 4, [1, 1], trainable=is_training,  ###第二条线产生预测框坐标，对预测框坐标进行预测
                                weights_initializer=initializer,
                                padding='VALID', activation_fn=None, scope='rpn_bbox_pred')
    if is_training:
      rois, roi_scores = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois") ###根据预测的类别和预测框坐标对porposa进行筛选，对前N个进行NMS,这里面nms只是找这些框,那些不是背景的得分高.把那些是背景概率高的框去掉.
      rpn_labels = self._anchor_target_layer(rpn_cls_score, "anchor")
      # Try to have a deterministic order for the computing graph, for reproducibility
      with tf.control_dependencies([rpn_labels]):
        rois, _ = self._proposal_target_layer(rois, roi_scores, "rpn_rois")
    else:
      if cfg.TEST.MODE == 'nms':
        rois, _ = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
      elif cfg.TEST.MODE == 'top':
        rois, _ = self._proposal_top_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
      else:
        raise NotImplementedError

    self._predictions["rpn_cls_score"] = rpn_cls_score
    self._predictions["rpn_cls_score_reshape"] = rpn_cls_score_reshape
    self._predictions["rpn_cls_prob"] = rpn_cls_prob
    self._predictions["rpn_cls_pred"] = rpn_cls_pred
    self._predictions["rpn_bbox_pred"] = rpn_bbox_pred
    self._predictions["rois"] = rois

    return rois

def _crop_pool_layer(self, bottom, rois, name): ####bottom为convert层卷积输出---也就是特征图， feat_stride为补偿乘积，用来求得原图的w,h.rois为选出的256个anchor的坐标,这些坐标是特征图上的坐标.
    '''
    结果就是在特征图上,把rois这些子图都扣出来.然后按照比例反映射到原始input_image里面的部分.
    类似感受眼这个东西.

    '''
    with tf.variable_scope(name) as scope:
      batch_ids = tf.squeeze(tf.slice(rois, [0, 0], [-1, 1], name="batch_id"), [1])
      # Get the normalized coordinates of bounding boxes
      bottom_shape = tf.shape(bottom)
      height = (tf.to_float(bottom_shape[1]) - 1.) * np.float32(self._feat_stride[0])
      width = (tf.to_float(bottom_shape[2]) - 1.) * np.float32(self._feat_stride[0])

      '''
      yinwei rois是针对原始图片的坐标.所以相对坐标是需要/htight or width
      '''
      x1 = tf.slice(rois, [0, 1], [-1, 1], name="x1") / width
      y1 = tf.slice(rois, [0, 2], [-1, 1], name="y1") / height
      x2 = tf.slice(rois, [0, 3], [-1, 1], name="x2") / width
      y2 = tf.slice(rois, [0, 4], [-1, 1], name="y2") / height###得到相对位置
      # Won't be back-propagated to rois anyway, but to save time

      '''
      因为bboxes.需要的是图片中子图的相对坐标位置,也就是4个百分比位置
      '''
      bboxes = tf.stop_gradient(tf.concat([y1, x1, y2, x2], axis=1))
      '''
      表示bboxes这个变量,不计算梯度.
      '''

      pre_pool_size = cfg.POOLING_SIZE * 2
      crops = tf.image.crop_and_resize(bottom, bboxes, tf.to_int32(batch_ids), [pre_pool_size, pre_pool_size], name="crops")##利用tensorflow的自带函数作用类似于ROI池化
      '''
      因为pre_pool_size = cfg.POOLING_SIZE * 2,所以下面再maxpool一下把图片缩小回去.
      '''
    return slim.max_pool2d(crops, [2, 2], padding='SAME')

import tensorflow as tf
# help(tf.image.crop_and_resize)

def _region_classification(self, fc7, is_training, initializer, initializer_bbox):
    cls_score = slim.fully_connected(fc7, self._num_classes,
                                       weights_initializer=initializer,
                                       trainable=is_training,
                                       activation_fn=None, scope='cls_score')
    cls_prob = self._softmax_layer(cls_score, "cls_prob")
    cls_pred = tf.argmax(cls_score, axis=1, name="cls_pred")
    bbox_pred = slim.fully_connected(fc7, self._num_classes * 4,
                                     weights_initializer=initializer_bbox,
                                     trainable=is_training,
                                     activation_fn=None, scope='bbox_pred')

    self._predictions["cls_score"] = cls_score
    self._predictions["cls_pred"] = cls_pred
    self._predictions["cls_prob"] = cls_prob
    self._predictions["bbox_pred"] = bbox_pred

    return cls_prob, bbox_pred