目标识别网络Faster RCNN：Pytorch源码分析（二）

锚框(anchors)产生：
感觉论文这块内容说的不太细致，只能深挖源码了~

from __future__ import print_function
# --------------------------------------------------------
# Faster R-CNN
# Copyright (c) 2015 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ross Girshick and Sean Bell
# --------------------------------------------------------

import numpy as np
import pdb

# Verify that we compute the same anchors as Shaoqing's matlab implementation:
#
#    >> load output/rpn_cachedir/faster_rcnn_VOC2007_ZF_stage1_rpn/anchors.mat
#    >> anchors
#
#    anchors =
#
#       -83   -39   100    56
#      -175   -87   192   104
#      -359  -183   376   200
#       -55   -55    72    72
#      -119  -119   136   136
#      -247  -247   264   264
#       -35   -79    52    96
#       -79  -167    96   184
#      -167  -343   184   360

#array([[ -83.,  -39.,  100.,   56.],
#       [-175.,  -87.,  192.,  104.],
#       [-359., -183.,  376.,  200.],
#       [ -55.,  -55.,   72.,   72.],
#       [-119., -119.,  136.,  136.],
#       [-247., -247.,  264.,  264.],
#       [ -35.,  -79.,   52.,   96.],
#       [ -79., -167.,   96.,  184.],
#       [-167., -343.,  184.,  360.]])

try:
    xrange          # Python 2
except NameError:
    xrange = range  # Python 3


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
    print(base_anchor)
    ratio_anchors = _ratio_enum(base_anchor, ratios)
    anchors = np.vstack([_scale_enum(ratio_anchors[i, :], scales)
                         for i in xrange(ratio_anchors.shape[0])])
    return anchors

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

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

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)
    print("haha", hs)
    anchors = _mkanchors(ws, hs, x_ctr, y_ctr)
    return anchors

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

if __name__ == '__main__':
    import time
    t = time.time()
    a = generate_anchors()
    print(time.time() - t)
    print(a)
    from IPython import embed; embed()

源码中除了第一个函数以外，其余函数都在第一个函数体调用过。我们注意第一个函数上来就有个base_anchor，这个可以理解为一个“基础框”，其他不同尺度和不同宽高比的锚框都是在“基础框”之上建立起来的，而且feature map同一个像素点的不同的9个锚框共用同一个几何中心点。base_anchor = [0, 0, 15, 15]，这指的是base_anchor的左下角坐标点(0,0)和右上角坐标点(15,15)。

generate_anchors函数的参数中有ratios=[0.5,1,2]和scales=[8,16,32]，ratios代表三种不同宽高比的锚框，scales代表三种不同尺度的锚框（即三种不同大小的锚框），这两者组合即可得到9种锚框。
_ratio_enum(anchor, ratios)函数中有调用_whctrs(anchor)，我们通过源码看一下_whctrs(anchor)函数是用来做什么的：

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

该函数将锚框(xmin, ymin, xmax, ymax)坐标形式转为(w,h,x_ctr,y_ctr)的形式，后者为锚框的宽、高以及中心点坐标表示形式。经转换，base_anchor的(0,0,15,15)变为(16,16,7.5,7.5)。为什么宽和高是16呢？因为这里计算的锚框的宽高是按照像素计算的，并不是根据数学中的长度来计量，并且base_anchor是从0像素点起始的。base_anchor的面积即为16x16=256，而且不同宽高比的锚框等于其面积比（即base_anchor/ratio得到的是宽高比为ratio的锚框），base_anchor即相当于宽高比为1的锚框，则其余情况得到的锚框面积即为256的1/2倍和2倍，size_ratios=[512,256,128]，然后
ws = np.round(np.sqrt(size_ratios))，hs = np.round(ws * ratios)（不太明白这里ws和hs的计算原理是啥~），得到ws=[23,16,11], hs=[12,16,22]，之后计算三种不同size_ratio下的锚框的坐标表示，是将(w,h,x_ctr,y_ctr)转为(xmin,ymin,xmax,ymax):

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

下面就该考虑不同尺度的锚框了：

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

手法类似，将上述的坐标(xmin,ymin,xmax,ymax)再次转为(w,h,x_ctr,y_ctr)的形式，w和h分别乘以尺度scales，即将宽高分别放大8倍、16倍、32倍，之后再次转为(xmin,ymin,xmax,ymax)的表达形式，即可得到不同尺度的锚框。