| by YoungTimes | No comments

Faster-RCNN中Anchor锚框生成

Anchor是Faster RCNN中的一个重要的概念,在对图像中的物体进行分类检测之前,先要生成一系列候选的检测框,以便于神经网络进行分类和识别。

图1-Faster RCNN中的锚框

1.什么是Anchor

论文中的描述如下:

An anchor is centered at the sliding window in question, and is associated with a scale and aspect ratio.

如图1所示,Anchor是以待检测位置为中心,以指定的大小和高宽比构成一组锚框。
假设Feature Map的宽度为W,宽度为H,在每个待检测的位置生成的锚框数目为K,根据滑动窗口的方法,生成总的锚框的数量是W * H * K。

2. Anchor的生成

在论文中,每个锚点有3种面积$Scale={{128}^2,{256}^{2},{512}^{2}}$和3种长宽比$AspectRatio={1:1, 1:2, 2:1}$,它们相互组合,每个Anchor生成9个锚框。

3. Anchor的代码实现

3.1 辅助函数

根据锚框得到其中心点(x_ctr,y_ctr)、宽度w、高度h。

def _whctrs(anchor):
  """
  Return width, height, x center, and y center for an anchor (window).
  """

  w = anchor[2] - anchor[0] + 1
  h = anchor[3] - anchor[1] + 1
  x_ctr = anchor[0] + 0.5 * (w - 1)
  y_ctr = anchor[1] + 0.5 * (h - 1)
  return w, h, x_ctr, y_ctr

根据宽度w、高度h、中心点(x_ctr,y_ctr)生成锚框。

def _mkanchors(ws, hs, x_ctr, y_ctr):
  """
  Given a vector of widths (ws) and heights (hs) around a center
  (x_ctr, y_ctr), output a set of anchors (windows).
  """

  ws = ws[:, np.newaxis]
  hs = hs[:, np.newaxis]
  anchors = np.hstack((x_ctr - 0.5 * (ws - 1),
                       y_ctr - 0.5 * (hs - 1),
                       x_ctr + 0.5 * (ws - 1),
                       y_ctr + 0.5 * (hs - 1)))
  return anchors

3.2 生成不同宽高比的锚框

def _ratio_enum(anchor, ratios):
  """
  Enumerate a set of anchors for each aspect ratio wrt an anchor.
  """

  w, h, x_ctr, y_ctr = _whctrs(anchor)
  size = w * h
  size_ratios = size / ratios
  ws = np.round(np.sqrt(size_ratios))
  hs = np.round(ws * ratios)
  anchors = _mkanchors(ws, hs, x_ctr, y_ctr)
  return anchors

对于同一个Anchor,不同的宽高比(Ratio)的面积是基本相同的。
记Anchor的面积为:$area=16*16$,宽高比:$ratio=w/h$,根据面积不变:

$$
area=w * h = h * ratio * h=ratio * h^2
$$
$$
h=\sqrt{area / ratio}
$$
$$
w=h* ratio=ratio * \sqrt{area / ratio}
$$
这也是上述代码的实现逻辑,代码中在根据ratio计算完w和h之后,进行了取整操作。在实际生成Anchors时,先定义一个大小为16 * 16的Base Anchor。

函数输入:
anchor=[0,0,15,15],ratios=[0.5, 1, 2]
函数输出:
[[-3.5, 2.0, 18.5, 13.0],
[0.0, 0.0, 15.0, 15.0],
[2.5, -3.0, 12.5, 18.0]]

3.生成不同比例的锚框

def _scale_enum(anchor, scales):
  """
  Enumerate a set of anchors for each scale wrt an anchor.
  """

  w, h, x_ctr, y_ctr = _whctrs(anchor)
  ws = w * scales
  hs = h * scales
  anchors = _mkanchors(ws, hs, x_ctr, y_ctr)
  return anchors

函数输入:
anchor=[-3.5, 2.0, 18.5, 13.0],scales=[8.0, 16.0, 24.0]
函数输出:
[[ -84.0, -40.0, 99.0, 55.0],
[-176.0, -88.0, 191.0, 103.0],
[-360.0, -184.0, 375.0 199.]]

4.锚框生成入口函数

def generate_anchors(base_size=16, ratios=[0.5, 1, 2],
                     scales=2 ** np.arange(3, 6)):
  """
  Generate anchor (reference) windows by enumerating aspect ratios X
  scales wrt a reference (0, 0, 15, 15) window.
  """

  base_anchor = np.array([1, 1, base_size, base_size]) - 1
  ratio_anchors = _ratio_enum(base_anchor, ratios)
  anchors = np.vstack([_scale_enum(ratio_anchors[i, :], scales)
                       for i in range(ratio_anchors.shape[0])])
  return anchors

以上代码生成9个锚框:
[[ -84. -40. 99. 55.]
[-176. -88. 191. 103.]
[-360. -184. 375. 199.]
[ -56. -56. 71. 71.]
[-120. -120. 135. 135.]
[-248. -248. 263. 263.]
[ -36. -80. 51. 95.]
[ -80. -168. 95. 183.]
[-168. -344. 183. 359.]]

四、锚框的效果

anchor效果

发表评论