r-cnn学习（六）：RPN及AnchorTargetLayer学习
RPN网络是faster与fast的主要区别，输入特征图，输出region proposals以及相应的分数。
# --------------------------------------------------------
# Faster R-CNN
# Copyright (c) 2015 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ross Girshick and Sean Bell
# --------------------------------------------------------
import os
import caffe
import yaml
from fast_rcnn.config import cfg
import numpy as np
import numpy.random as npr
from generate_anchors import generate_anchors
from utils.cython_bbox import bbox_overlaps
from fast_rcnn.bbox_transform import bbox_transform
DEBUG = False
class AnchorTargetLayer(caffe.Layer):
"""
Assign anchors to ground-truth targets. Produces anchor classification
labels and bounding-box regression targets.
"""
#生成anchors，reshap输出
def setup(self, bottom, top):
layer_params = yaml.load(self.param_str_)
anchor_scales = layer_params.get('scales', (8, 16, 32))
self._anchors = generate_anchors(scales=np.array(anchor_scales))#九个anchor的w h x_cstr y_cstr，对原始的wh做横向纵向变化，并放大缩小得到九个
self._num_anchors = self._anchors.shape[0]<span style="font-family: Arial, Helvetica, sans-serif;">#anchor的个数</span>
self._feat_stride = layer_params['feat_stride']#网络中参数16 （feature map为原图大小的1/16，1000*600->60*40）
if DEBUG:
print 'anchors:'
print self._anchors
print 'anchor shapes:'
print np.hstack((
self._anchors[:, 2::4] - self._anchors[:, 0::4],
self._anchors[:, 3::4] - self._anchors[:, 1::4],
))
self._counts = cfg.EPS
self._sums = np.zeros((1, 4))
self._squared_sums = np.zeros((1, 4))
self._fg_sum = 0
self._bg_sum = 0
self._count = 0
# allow boxes to sit over the edge by a small amount
self._allowed_border = layer_params.get('allowed_border', 0)
#bottom 长度为4；bottom[0],map；bottom[1],boxes,labels;bottom[2],im_fo;bottom[3],图片数据
height, width = bottom[0].data.shape[-2:]
if DEBUG:
print 'AnchorTargetLayer: height', height, 'width', width
A = self._num_anchors#anchor的个数
# labels
top[0].reshape(1, 1, A * height, width)
# bbox_targets
top[1].reshape(1, A * 4, height, width)
# bbox_inside_weights
top[2].reshape(1, A * 4, height, width)
# bbox_outside_weights
top[3].reshape(1, A * 4, height, width)
#每个位置生成9个anchor，通过GT overlap过滤掉一部分anchors
def forward(self, bottom, top):
# Algorithm:
#
# for each (H, W) location i
# generate 9 anchor boxes centered on cell i
# apply predicted bbox deltas at cell i to each of the 9 anchors
# filter out-of-image anchors
# measure GT overlap
assert bottom[0].data.shape[0] == 1, \
'Only single item batches are supported'
#取得相应的anchors的h,w以及gt_box的位置，label
# map of shape (..., H, W)
height, width = bottom[0].data.shape[-2:]
# GT boxes (x1, y1, x2, y2, label)
gt_boxes = bottom[1].data#gt_boxes:长度不定
# im_info
im_info = bottom[2].data[0, :]
if DEBUG:
print ''
print 'im_size: ({}, {})'.format(im_info[0], im_info[1])
print 'scale: {}'.format(im_info[2])
print 'height, width: ({}, {})'.format(height, width)
print 'rpn: gt_boxes.shape', gt_boxes.shape
print 'rpn: gt_boxes', gt_boxes
#算出box的偏移量
# 1. Generate proposals from bbox deltas and shifted anchors
shift_x = np.arange(0, width) * self._feat_stride
shift_y = np.arange(0, height) * self._feat_stride
shift_x, shift_y = np.meshgrid(shift_x, shift_y)
shifts = np.vstack((shift_x.ravel(), shift_y.ravel(),
shift_x.ravel(), shift_y.ravel())).transpose()
# add A anchors (1, A, 4) to 根据偏移量移动anchors
# cell K shifts (K, 1, 4) to get
# shift anchors (K, A, 4)
# reshape to (K*A, 4) shifted anchors
A = self._num_anchors
K = shifts.shape[0]
all_anchors = (self._anchors.reshape((1, A, 4)) +
shifts.reshape((1, K, 4)).transpose((1, 0, 2)))
all_anchors = all_anchors.reshape((K * A, 4))
total_anchors = int(K * A)#K*A，所有anchors个数，包括越界的
#K: width*height
#A: 9
# only keep anchors inside the image
inds_inside = np.where(
(all_anchors[:, 0] >= -self._allowed_border) &
(all_anchors[:, 1] >= -self._allowed_border) &
(all_anchors[:, 2] < im_info[1] + self._allowed_border) & # width
(all_anchors[:, 3] < im_info[0] + self._allowed_border) # height
)[0]#没有过界的anchors索引
if DEBUG:
print 'total_anchors', total_anchors
print 'inds_inside', len(inds_inside)
# keep only inside anchors
anchors = all_anchors[inds_inside, :]#没有过界的anchors
if DEBUG:
print 'anchors.shape', anchors.shape
# label: 1 is positive, 0 is negative, -1 is dont care
labels = np.empty((len(inds_inside), ), dtype=np.float32)
labels.fill(-1)
# overlaps between the anchors and the gt boxes
# overlaps (ex, gt)
overlaps = bbox_overlaps( #返回大小连续的overlaps，等同于排序
np.ascontiguousarray(anchors, dtype=np.float),
np.ascontiguousarray(gt_boxes, dtype=np.float))
#找到某个box与所有gt_box最大的overlaps
argmax_overlaps = overlaps.argmax(axis=1)#overlaps每行最大值索引
max_overlaps = overlaps[np.arange(len(inds_inside)), argmax_overlaps]#最大的overlaps值
#找到某gt_box与所有box最大的overlaps
gt_argmax_overlaps = overlaps.argmax(axis=0) #overlaps每列中最大值索引
gt_max_overlaps = overlaps[gt_argmax_overlaps,#其对应的overlaps值
np.arange(overlaps.shape[1])]
gt_argmax_overlaps = np.where(overlaps == gt_max_overlaps)[0]
if not cfg.TRAIN.RPN_CLOBBER_POSITIVES:
# assign bg labels first so that positive labels can clobber them
labels[max_overlaps < cfg.TRAIN.RPN_NEGATIVE_OVERLAP] = 0
//对于某个gt，overlap最大的anchor为1
# fg label: for each gt, anchor with highest overlap
labels[gt_argmax_overlaps] = 1
//对于某个anchor，其overlap超过某值为1
# fg label: above threshold IOU
labels[max_overlaps >= cfg.TRAIN.RPN_POSITIVE_OVERLAP] = 1
if cfg.TRAIN.RPN_CLOBBER_POSITIVES:
# assign bg labels last so that negative labels can clobber positives
labels[max_overlaps < cfg.TRAIN.RPN_NEGATIVE_OVERLAP] = 0
# subsample positive labels if we have too many 如果正样本较多，降采样
num_fg = int(cfg.TRAIN.RPN_FG_FRACTION * cfg.TRAIN.RPN_BATCHSIZE) //正样本数量
fg_inds = np.where(labels == 1)[0]
if len(fg_inds) > num_fg:
disable_inds = npr.choice(
fg_inds, size=(len(fg_inds) - num_fg), replace=False)
labels[disable_inds] = -1 //多余正样本被随机标为负样本（这样真的好吗？）
# subsample negative labels if we have too many 同样处理负样本
num_bg = cfg.TRAIN.RPN_BATCHSIZE - np.sum(labels == 1)
bg_inds = np.where(labels == 0)[0]
if len(bg_inds) > num_bg:
disable_inds = npr.choice(
bg_inds, size=(len(bg_inds) - num_bg), replace=False)
labels[disable_inds] = -1 //仍然标为负？
#print "was %s inds, disabling %s, now %s inds" % (
#len(bg_inds), len(disable_inds), np.sum(labels == 0))
#保留最大overlaps的anchors，其他为0（非极大值抑制？）
bbox_targets = np.zeros((len(inds_inside), 4), dtype=np.float32)
bbox_targets = _compute_targets(anchors, gt_boxes[argmax_overlaps, :]) ##
#正样本inside_weights为1，其余为0（等同于论文中的pi*）
bbox_inside_weights = np.zeros((len(inds_inside), 4), dtype=np.float32)
bbox_inside_weights[labels == 1, :] = np.array(cfg.TRAIN.RPN_BBOX_INSIDE_WEIGHTS)
bbox_outside_weights = np.zeros((len(inds_inside), 4), dtype=np.float32)
#对样本权重进行归一化
if cfg.TRAIN.RPN_POSITIVE_WEIGHT < 0: # uniform weighting of examples (given non-uniform sampling)
num_examples = np.sum(labels >= 0)
positive_weights = np.ones((1, 4)) * 1.0 / num_examples
negative_weights = np.ones((1, 4)) * 1.0 / num_examples
else:
assert ((cfg.TRAIN.RPN_POSITIVE_WEIGHT > 0) &
(cfg.TRAIN.RPN_POSITIVE_WEIGHT < 1))
positive_weights = (cfg.TRAIN.RPN_POSITIVE_WEIGHT /
np.sum(labels == 1))
negative_weights = ((1.0 - cfg.TRAIN.RPN_POSITIVE_WEIGHT) /
np.sum(labels == 0))
bbox_outside_weights[labels == 1, :] = positive_weights
bbox_outside_weights[labels == 0, :] = negative_weights
if DEBUG:
self._sums += bbox_targets[labels == 1, :].sum(axis=0)
self._squared_sums += (bbox_targets[labels == 1, :] ** 2).sum(axis=0)
self._counts += np.sum(labels == 1)
means = self._sums / self._counts
stds = np.sqrt(self._squared_sums / self._counts - means ** 2)
print 'means:'
print means
print 'stdevs:'
print stds
# map up to original set of anchors 对total_anchors的其他box，weights及label进行填充
labels = _unmap(labels, total_anchors, inds_inside, fill=-1)
bbox_targets = _unmap(bbox_targets, total_anchors, inds_inside, fill=0)
bbox_inside_weights = _unmap(bbox_inside_weights, total_anchors, inds_inside, fill=0)
bbox_outside_weights = _unmap(bbox_outside_weights, total_anchors, inds_inside, fill=0)
if DEBUG:
print 'rpn: max max_overlap', np.max(max_overlaps)
print 'rpn: num_positive', np.sum(labels == 1)
print 'rpn: num_negative', np.sum(labels == 0)
self._fg_sum += np.sum(labels == 1)
self._bg_sum += np.sum(labels == 0)
self._count += 1
print 'rpn: num_positive avg', self._fg_sum / self._count
print 'rpn: num_negative avg', self._bg_sum / self._count
# labels 输出标签、box、inside_weights、outside_weights
labels = labels.reshape((1, height, width, A)).transpose(0, 3, 1, 2)
labels = labels.reshape((1, 1, A * height, width))
top[0].reshape(*labels.shape)
top[0].data[...] = labels
# bbox_targets
bbox_targets = bbox_targets \
.reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2)
top[1].reshape(*bbox_targets.shape)
top[1].data[...] = bbox_targets
# bbox_inside_weights
bbox_inside_weights = bbox_inside_weights \
.reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2)
assert bbox_inside_weights.shape[2] == height
assert bbox_inside_weights.shape[3] == width
top[2].reshape(*bbox_inside_weights.shape)
top[2].data[...] = bbox_inside_weights
# bbox_outside_weights
bbox_outside_weights = bbox_outside_weights \
.reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2)
assert bbox_outside_weights.shape[2] == height
assert bbox_outside_weights.shape[3] == width
top[3].reshape(*bbox_outside_weights.shape)
top[3].data[...] = bbox_outside_weights
def backward(self, top, propagate_down, bottom):
"""This layer does not propagate gradients."""
pass
def reshape(self, bottom, top):
"""Reshaping happens during the call to forward."""
pass
def _unmap(data, count, inds, fill=0):
""" Unmap a subset of item (data) back to the original set of items (of
size count) """ #对于total_anchors，保留设定的label，其余填为fill
if len(data.shape) == 1:
ret = np.empty((count, ), dtype=np.float32)
ret.fill(fill)
ret[inds] = data
else:
ret = np.empty((count, ) + data.shape[1:], dtype=np.float32)
ret.fill(fill)
ret[inds, :] = data
return ret
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32, copy=False)
算偏移量时涉及到的公式：
这段代码主要生成anchors，算出anchors的偏移量，并根据与gt的overlaps，进行NMS及排序，赋予其相应的标签。
其中generate_anchors.py的源码如下。这段代码生成不同宽高比(1:2,1:1,2:1)、不同尺度（8 16 32）的anchors：
<span style="font-size:24px;">#功能描述：生成多尺度、多宽高比的anchors。
# 尺度为：128,256,512; 宽高比为：1:2,1:1,2:1
import numpy as np #提供矩阵运算功能的库
#生成anchors总函数：ratios为一个列表，表示宽高比为：1:2,1:1,2:1
#2**x表示:2^x，scales:[2^3 2^4 2^5],即:[8 16 32]
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 #新建一个数组：base_anchor:[0 0 15 15]
ratio_anchors = _ratio_enum(base_anchor, ratios) #枚举各种宽高比
anchors = np.vstack([_scale_enum(ratio_anchors[i, :], scales) #枚举各种尺度，vstack:竖向合并数组
for i in xrange(ratio_anchors.shape[0])]) #shape[0]:读取矩阵第一维长度，其值为3
return anchors
#用于返回width,height,(x,y)中心坐标(对于一个anchor窗口)
def _whctrs(anchor):
"""
Return width, height, x center, and y center for an anchor (window).
"""
#anchor:存储了窗口左上角，右下角的坐标
w = anchor[2] - anchor[0] + 1
h = anchor[3] - anchor[1] + 1
x_ctr = anchor[0] + 0.5 * (w - 1) #anchor中心点坐标
y_ctr = anchor[1] + 0.5 * (h - 1)
return w, h, x_ctr, y_ctr
#给定一组宽高向量，输出各个anchor，即预测窗口，**输出anchor的面积相等，只是宽高比不同**
def _mkanchors(ws, hs, x_ctr, y_ctr):
#ws:[23 16 11]，hs:[12 16 22],ws和hs一一对应。
"""
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] #newaxis:将数组转置
hs = hs[:, np.newaxis]
anchors = np.hstack((x_ctr - 0.5 * (ws - 1), #hstack、vstack:合并数组
y_ctr - 0.5 * (hs - 1), #anchor：[[-3.5 2 18.5 13]
x_ctr + 0.5 * (ws - 1), # [0 0 15 15]
y_ctr + 0.5 * (hs - 1))) # [2.5 -3 12.5 18]]
return anchors
#枚举一个anchor的各种宽高比，anchor[0 0 15 15],ratios[0.5,1,2]
def _ratio_enum(anchor, ratios):
""" 列举关于一个anchor的三种宽高比 1:2,1:1,2:1
Enumerate a set of anchors for each aspect ratio wrt an anchor.
"""
w, h, x_ctr, y_ctr = _whctrs(anchor) #返回宽高和中心坐标，w:16,h:16,x_ctr:7.5,y_ctr:7.5
size = w * h #size:16*16=256
size_ratios = size / ratios #256/ratios[0.5,1,2]=[512,256,128]
#round()方法返回x的四舍五入的数字，sqrt()方法返回数字x的平方根
ws = np.round(np.sqrt(size_ratios)) #ws:[23 16 11]
hs = np.round(ws * ratios) #hs:[12 16 22],ws和hs一一对应。as:23&12
anchors = _mkanchors(ws, hs, x_ctr, y_ctr) #给定一组宽高向量，输出各个预测窗口
return anchors
#枚举一个anchor的各种尺度，以anchor[0 0 15 15]为例,scales[8 16 32]
def _scale_enum(anchor, scales):
""" 列举关于一个anchor的三种尺度 128*128,256*256,512*512
Enumerate a set of anchors for each scale wrt an anchor.
"""
w, h, x_ctr, y_ctr = _whctrs(anchor) #返回宽高和中心坐标，w:16,h:16,x_ctr:7.5,y_ctr:7.5
ws = w * scales #[128 256 512]
hs = h * scales #[128 256 512]
anchors = _mkanchors(ws, hs, x_ctr, y_ctr) #[[-56 -56 71 71] [-120 -120 135 135] [-248 -248 263 263]]
return anchors
if __name__ == '__main__': #主函数
import time
t = time.time()
a = generate_anchors() #生成anchor（窗口）
print time.time() - t #显示时间
print a
from IPython import embed; embed()
</span>
参考：http://blog.csdn.net/u010668907/article/details/51942481
http://blog.csdn.net/xzzppp/article/details/52317863
